AN INTRODUCTION TO

HYDRODYNAMICS
AND WATER WAVES
HYDR ODYN AMIC A
SIVE
DE VIRIBUS ET MOTIBUS FLUIDORUM
COMMENTA RII

'Remember, when discoursing about water, to induce

first experience, then reason.'
- Leonardo da Vinci
 AN INTRODUCTION TO

HYDRODYNAMICS
AND WATER WAVES

Bernard Le Mehaute

SPRINGER SCIENCE+ BUSINESS MEDIA, LLC 1976

Bernard Le Mehaute
Senior Vice President
Tetra Tech, Inc.
630 North Rosemead Boulevard
Pasadena, California 91107

This volume has been totally revised from an

earlier version published by the U.S. Government
Printing Office as ESSA Technical Reports
ERL 118-POL 3-1 and 3-2.

1976 by Springer Science+Business Media New York

Originally published by Springer-Verlag New York in 1976
Softcover reprint of the hardcover 1st edition 1976

All rights reserved.

No part of this book may be translated or
reproduced in any form without written permission
from Springer Science+Business Media, LLC .

Library of Congress Cataloging in Publication Data

Le Mehaute, Bernard, 1927-

An introduction to hydrodynamics and water waves.
Includes bibliographical references and index.
I. Hydrodynamics. 2. Water waves. I. Title.
QA9ll.L39 532'.5 75-12754

ISBN 978-3-642-85569-6 ISBN 978-3-642-85567-2 (eBook)

DOI 10.1007/978-3-642-85567-2
Additional material to this book can be downloaded from http://extras.springer.com
 Hydrodynamics is the science which deals with the v
Preface motion of liquid in the macroscopic sense. It is essentially
a field which is regarded as applied mathematics because it
deals with the mathematical treatments of basic equations
for a fluid continuum obtained on a purely Newtonian
basis. It is also the foundation of hydraulics, which, as an
art, has to compromise with the rigorous mathematical
treatments because of nonlinear effects, inherent instability,
turbulence, and the complexity of" boundary conditions"
encountered in engineering practice. Therefore, this book
can be considered as the text for a course in basic hydro-
dynamics, as well as for a course in the fundamentals of
hydraulic and related engineering disciplines.
In the first case, the students learn how to make use of
their mathematical knowledge in a field of physics particu-
larly suitable to mathematical treatments. Since they may
have some difficulty in representing a physical phenomenon
by a mathematical model, a great emphasis has been given
to the physical concepts of hydrodynamics. For students
with an undergraduate training in engineering, the diffi-
culty may be a lack of appropriate mathematical tools.
Their first contact with hydraulics has been on an essen-
tially practical basis. They may be discouraged in attempt-
ing the study of such books as Hydrodynamics, by Lamb,
which remains the bible of hydrodynamicists. Hence,
mathematical intricacies have been introduced slowly and
progressively. Also, the emphasis on the physical approach
has made it possible to avoid mathematical abstractions so
that a concrete support may be given to equations.
Finally, the author has tried to make this book self-
contained in the sense that a practicing engineer who wants
to improve his theoretical background can study hydro-
dynamics by himself without attending lectures. Too often
articles in scientific journals present some discouraging
aspects to practicing engineers and the most valuable
messages can only reach a few specialists. It is felt that the
learning of some basic theories will help hydraulic engi-
neers to keep abreast of and participate in new develop-
ments proposed by theorists.
Considering that a good assimilation of the basis is
essential before further study, great care has been taken to
vi develop a clear understanding, both mathematically and motion and dynamical oceanography. It is hoped also that
physically, of the fundamental concepts of theoretical hy- it will instill in engineering students the desire for further
draulics. The introduction of mathematical simplifications study in hydrodynamics and mathematics. It is also hoped
and assumptions, often based on physical considera- that the book will be of great help to students in hydraulics,
tions, has also been developed by examples. The mathe- civil and coastal engineering, naval architecture, as well as
matical difficulties have been cleared up by introducing in physical oceanography, marine geology, and sedimentol-
them progressively and by developing all the intermediate ogy, who want to learn or revise one of the theoretical
calculations. Also, all the abstract concepts of theoretical aspect of their future profession.
hydraulics have been explained as concretely as possible by
use of examples. It will appear that the first chapter is the
easiest to understand, and it is assumed that the mathe- ACKNOWLEDGMENTS
matical background increases as the student progresses
toward the end of the book. However, it is taken for granted The author wishes to express his deep gratitude to
that the student already has some notion of elementary Nicholas Boratynski, President of Tetra Tech, Inc. Without
hydraulics. his encouragement and support, these lecture notes would
Finally, the succession of the various chapters have been not have been revised and published.
chosen in order to build up a structure as logical and as Many valuable suggestions and contributions are credited
deductive as possible in order to avoid that the various to the senior engineering and scientific staff of Tetra Tech.
subjects appear as a succession of different mathematical The author would also like to acknowledge Dr. Viviane
recipes rather than as a unique and logical subject. Rupert of the Lawrence Livermore Laboratory for her
Part One deals with the establishment of the fundamental help in editing the book.
differential equations governing the flow motion in all
possible cases. The possible approximations are also Bernard Le Mehaute
indicated. Part Two deals with general methods of integra- Pasadena, California
tions and the mathematical treatments of these equations.
Integrations of general interest, and integrations in some
typical particular cases are presented. Part Three is devoted
to water wave theories, as one of the most important topics
of hydrodynamics.
It is pointed out that the emphasis of the book is on water
waves. Therefore the treatment of motion of compressible
fluids has been judged beyond the scope of this book, with
a few exceptions. Also, almost all the calculations are
presented in a Cartesian (or cylindrical) system of coordi-
nates. Vectorial and tensorial operations have been mini-
mized in order to reduce the necessary mathematical
background. However, vectorial and tensorial notations
are slowly introduced for sake of recognition in the
literature.
It is hoped that this book will entice students gifted in
mathematics to apply their capabilities to the study of fluid
Contents PART ONE vii
Establishing the Basic Equations that Govern
Flow Motion
1. Basic Concepts and Principles 3
2. Motions of Fluid Elements; Rotational and
Irrotational Flow 16
3. The Continuity Principle 31
4. Inertia Forces 37
5. Applied Forces 51
6. Forms of the Momentum Equation: Equations of
Euler and Navier-Stokes- 59
7. Turbulence: Mean and Fluctuating Components of
Motion 71
8. Turbulence Effects: Modern Theories 80
9. Flow in a Porous Medium: Law of Darcy 89

PART TWO

Some Mathematical Treatments of the Basic

Equations
10. The Bernoulli Equation 101
11. Flow Pattern, Stream Function, Potential
Function 116
12. The Momentum Theorem and Its Applications 137
13. The Boundary Layer, Flow in Pipes, Drag and
Added Mass 154
14. Open-Channel Hydraulics 182

PART THREE

Water Wave Theories

15. An Introduction to Water Waves 197
16. Linear Small Amplitude Wave Theories 212
 Contents

vm 17. Finite Amplitude Waves 239

18. The Long Wave Theory 256

Appendix A: Wave Motion as a Random

Process 273
Appendix B: Similitude and Scale Model
Technology 286
Notation 301
Answers to Selected Problems 305
Index 317
 PART ONE

Establishing the Basic

Equations that Govern
Flow Motion
Chapter
3
I 1-1 Basic Concepts of Hydrodynamics
1-1.1 Definition of an Elementary
Particle of Fluid

Basic Concepts Studies of theoretical fluid mechanics are based on the

concept of an elementary mass or particle of fluid. This

and Principles particle has no well-defined existence. It may be considered

as a corpus alienum, a foreign body in the mechanics of a
continuum. It is an aid toward the understanding of the
physical meaning of the differential equations governing
the flow motion.
Just as the fundamental concepts of the theoretical
mechanics of solid matter are based on the mechanics of a
so-called "material point," the basis of theoretical fluid
mechanics rests on the mechanics of an elementary mass
of fluid. Such an elementary mass of fluid, in common with
the material point in the kinematics of a solid body, is
assumed to be either infinitely small or small enough that
all parts of the element can be considered to have the same
velocity of translation V and the same density p. This
elementary fluid particle is assumed to be homogeneous,
isotropic, and continuous in the macroscopic sense. The
molecular pattern and the molecular and Brownian motions
within the particle, a subject dealt with in the kinetic theory
of fluids, are not taken into account.
1-1.2 Theoretical Approach
The laws of mechanics of a solid body system (a rotating
disk, for example) are obtained by the integration of the
laws of mechanics for a "material point" with respect to the
area or the volume of the system under consideration.
Similarly, the laws of fluid mechanics used in engineering
practice are obtained by integration-exact or approximate
-of the laws governing the behavior of a fluid particle
along a line or throughout an area or a volume. Hence,
studies in hydrodynamics may be divided into two different
parts.

1-1.2.1 The first part consists of establishing the general

differential equations which govern the motion of an
 Part 1 : Establishing the Basic Equations that Govern Flow Motion

4 elementary particle of fluid. The fluid rna y be assumed either

perfect (without friction forces) or real. In the latter case, 1-2 Streamline, Path, Streakline, and
the flow may be either laminar or turbulent. Stream Tube
1-1.2.2 The second step involves the study of different 1-2.1 Notation
mathematical methods used to integrate these basic Consider the point A(x,y,z) in a Cartesian system of
differential equations. Practical general relationships, such coordinates. The axes OX, OY, OZ are mutually perpen-
as the well-known Bernoulli equation, may thereby be pendicular (see Fig. 1-1). Consider an infinitely small
deduced. Solutions, valid for special cases, can also be rectangular element of fluid with the point as a corner.
obtained by direct integration. The edges of this element are dx, dy, dz. Its volume is
dx dy dz and its weight is 05 dx dy dz or pg dx dy dz. 05 is the
1-1.3 Relations between Fluid Particles:
specific weight and g is the acceleration due to gravity.
Friction Forces
The pressure at point A is a scalar quantity which is
In a solid material, points in a system (on a disk, for completely specified by its magnitude. The pressure is
example) do not change their relative position (except for always exerted perpendicular to the considered surface (see
elastic deformations which are described by well-defined Section 5-3.1). The corresponding force is a vector quantity,
laws). On the other hand, fluid particles may be deformed which is specified by its magnitude and direction. The
and each particle may have a particular motion which magnitude of the pressure p is a function of the space
differs quite markedly from the motion of other particles. coordinates of A and timet; i.e., p = f(x,y,z,t). Its direction
The forces exerted between fluid particles are the pressure is normal to the area on which the pressure is exerted. The
forces and the friction forces.
The friction force per unit area in a given direction, called
the shear stress r, is assumed to be either zero ("ideal" Figure 1-1 Notation in Cartesian coordinates.
or perfect fluid), or proportional to the coefficient of
z
viscosity 11 (viscous fluid). The shear stress r is a scalar.
The set of shear stresses at a point constitutes a tensor.
I
The significance of this statement is developed in Chapter 5. I
For now it is sufficient to know that the shearing stress, I
I
at any point of a plane parallel to a unidirectional flow is I
I
v
'=
dV ~--/
11-
dn !
I
\If--
where n is the perpendicular direction to the flow moving 1 I
with velocity V. I I
I I
Hydrodynamics is primarily concerned with a I I
X

"Newtonian fluid," that is, its viscous stress tensor depends I I I

I I I
linearly, isotropically, and covariantly (Chapter 5) on the
rate of strain or derivatives of the velocity components. !v ----~~ /
It does not deal with "plastic" fluids where the coefficient 11 I I I
is replaced by a function of the intensity or duration of the ------------
shear. y
 Chapter 1 : Basic Concepts and Principles

gradient of p (grad p or Vp), its derivative with respect to

--
VELOCITY VECTORS
5
/

---
space, is also a vector quantity. The components of grad p / /
along the three coordinate axes OX, OY, OZ, are given by
/ / .& STREAMLINE
the derivative of p with respect to x, y, z, respectively; i.e.,
/ ~
8pj8x, 8pj8y, 8pj8z. / /

The velocity of fluid particles at A is V. The components I ~ / /

ofV along the three Cartesian coordinate axes OX, OY, OZ,
are u, v, and w, respectively. If i, j, k are unit vectors along
the axes OX, OY, OZ respectively, then: V = iu + jv + kw.
Since the system of reference is rectangular, the magnitude
of the velocity is given by V = [u 2 + v2 + w2 2 V is a r1
scalar quantity and therefore completely defined by its
magnitude, like the pressure p. V is a vector quantity and is
specified by its direction and magnitude. Since V and its
components u, v, and w are functions of the space co-
ordinates of A and the timet, they can be written in the form
V(x,y,z,t).
1-2.2 Definitions
1-2.2.1 The displacement dS of a fluid particle is defined
by the vector equation, dS = V dt, which is valid for both
magnitude and direction. This equation may be written
more specifically in terms of the displacements in each of
the three Cartesian coordinate directions as follows:
dx = u dt
dy = v dt
dz = w dt
I ~ / /
I
I I /
Figure 1-2 Streamlines observed by short-exposure
photography of various particles.
This is the mathematical definition of a streamline. These
equalities express the fact that the velocity is tangential to
the displacement of the particle at time t 0 . Figure 1-3
illustrates this fact in the case of a two-dimensional motion.
In this case dxju = dyjv, which implies v dx - u dy = 0.
Streamlines do not cross, except at point of theoretically
infinite velocity (see Figs. 11-6 and 11-7) and at stagnation
and separation points of a body where the velocity is zero.
Fixed solid boundaries and steady free surfaces are stream-
lines. Moving boundaries, such as propeller blades, and
unsteady free surfaces are not streamlines.
1-2.2.3 The path of a specific particle of the fluid is
defined by its position as a function of time. It may be

1-2.2.2 A streamline is defined as a line which is tangential Figure 1-3 Definition of a streamline in a
at every point to the velocity vector at a given time t 0 . two-dimensional motion.
A device for visualizing streamlines is to imagine a number y
of small bright particles distributed at random in the fluid,
and then to photograph them with a short exposure v
(Fig. 1-2). Every particle photographs as a small line
segment. Each line which is drawn tangentially to these
small segments is a streamline.
v 1
dy
At time t 0 , the equations dx = u dt, dy = v dt, and
dz = w dt become:
dx dy dz
0~--~=d=x::~:---~x
u(x,y,z,t 0 ) v(x,y,z,t 0 ) w(x,y,z,t 0 ) u
 Part 1 : Establishing the Basic Equations that Govern Flow Motion
6 determined by photographing a bright particle with a
long exposure. The path line is tangential to the streamline
at a given time t 0 . However, the time has to be included
as a variable for defining a path. Hence, the path lines are
defined mathematically as
dx dy dz = dt
u(x,y,z,t) v(x,y,z,t) w(x,y,z,t)
1-2.2.4 A streakline is given by an instantaneous shot
photographing a number of small bright particles in
suspension which were introduced into the fluid at the
same point at regular intervals of time (Fig. 1-4).
1-2.2.5 An elementary flow channel bounded by an
infinite number of streamlines crossing a closed curve
is known as a stream tube (Fig. 1-5).
1-2.3 Steady and Unsteady Flow
1-2.3.1 For steady flows defined by time-independent
quantities, streamlines, streaklines, and particle paths
are identical. However, for unsteady or time-dependent
flows, these lines are different and a clear understanding
of their generation is necessary to properly interpret the
results of a given experiment. For example, if dye is injected
at a given point of a fluid flow, the dye pattern will be a
Figure 1-4 Streakline obtained by instantaneous
photography of various particles
coming from the same point.
Figure 1-5 Stream tube.
streakline; if the successive location of a neutrally buoyant
small ball are determined, a particle path can be traced;
finally, if a large number of short threads are attached to a
body, the instantaneous direction of these threads will
yield a streamline pattern. All these methods are commonly
used in fluid flow experimental studies.
Streamlines, paths, streaklines, and stream tubes are
different in unsteady flow, that is, flow changing with
respect to time. Turbulent flow is always an unsteady flow;
however, it will be seen in that case that the mean motion
with respect to time of a turbulent flow may be considered
as steady. Then streamlines, paths, and streaklines of the
mean motion are the same (see Chapter 7). Figures 1-6 and
1-7 illustrate these definitions in some cases of unsteady
motion.
1-2.3.2 In some cases of unsteady flow (a body moving
at constant velocity in a still fluid, a steady wave profile
such as those due to a periodic wave or a solitary wave) it is
possible to transform an unsteady motion into a steady
motion relative to a coordinate system which moves with
the body or the wave velocity. The construction of a steady
pattern is then obtained by subtracting the velocity of the
body from the velocity of the fluid. This is the Galilean
transformation. Steady streamlines can then be defined
relative to a moving observer who travels with the body or
with the wave (see Fig. 1-8).

Figure 1-6
Periodic gravity wave in deep water.
1-3 Methods of Study
The motion of a fluid can be studied either by the method
of Lagrange or the method of Euler.
1-3.1 Lagrangian Method
The Lagrangian method may be used to answer the
question: What occurs to a given particle of fluid as it
moves along its own path? This method consists offollowing
the fluid particles during the course of time and giving the
paths, velocities, and pressures in terms of the original
position of the particles and the time elapsed since the
particles occupied their original position. In the case of a
WIND
Figure 1-7
Smoke in the wind.
Chapter 1 : Basic Concepts and Principles
WAVE TRAVEL 7
compressible fluid, densities and temperatures are also
given in terms of the original position and the elapsed time.
If the initial position of a given particle at time t 0 is
x 0 , y 0 , z0 , a Lagrangian system of equations gives the
position x, y, z, at the instant t as:
X= F 1(x 0 ,y0 ,z 0 ,t- t 0 )
y = F 2 (x 0 ,y0 ,z0 ,t- t 0 )
z = F 3 (x 0 ,y0 ,z 0 ,t - t 0 )
In practice this method is seldom used in hydrodynamics.
Lagrangian coordinates are, however, often used in theories
relative to periodical gravity waves. The velocity and
... SMOKE
 Part I : Establishing the Basic Equations that Govern Flow Motion

8 Figure 1-8 (Top) Streamlines, paths, streaklines for a acceleration components at point (x 0 ,y0 ,z 0 ) are then
steady flow around a .fixed body. obtained by a simple partial differentiation with respect
(Middle) Streamlines, paths for an unsteady to time, such that
flow around a body moving at constant
velocity in a stillfluid.
(Bottom) Vectorial relationship between the
U =
OX
7ft
I
xo,yo,zo
two kinds of motion: Galilean
transformation.

oz I
W = Ot Xo,yo,zo

v Similarly, the acceleralion components are o 2 xjot 2 , o2 yjot 2 ,

o 2 z/ot 2

1-3.2 Eulerian Method

The Eulerian method may be used to answer the
question: What occurs at a given point in a space occupied
by a fluid in motion? This is the most frequent form of
problem encountered in hydrodynamics. This method gives,
at a given point A(x,y,z), the velocity V(u,v,w) and the
pressure p (and, in the case of a compressible fluid, density
and temperature) as functions of timet. Since

V = F(x,y,z,t)
then
u f 1(x,y,z,t)
=

v = f 2 (x,y,z,t)
w = f 3 (x,y,z,t)
and
p = F 1(x,y,z,t)

The Eulerian system of equations is found by a total

differentiation of u, v, and w with respect to t and by con-
sideration of the pressure components. In the following
example the Eulerian system of coordinates is used.
 Chapter 1 : Basic Concepts and Principles

where t is a variable. If it can be assumed that x and z differ 9

little from some given values x 0 and z0 , the differential
1-3.3 An Example of Flow Pattern
equation, to a first approximation becomes:
Let us consider an Eulerian system of coordinates where
H
the two-dimensional wave motion is represented by the dx = k 2 emzo cos (kt - mx 0 ) dt
velocity components:
Hence,
dx H
u= f 1(x,z,t) = dt = 2 kemz cos (kt - mx)
X - X; = ~ emzo sin (kt - mXo)
w= f 3 (x,z,t) = dz =- H kemz sin (kt - mx)
dt 2 (z - z;) is found by a similar procedure. It is

The equations for the streamlines are obtained from the

Z - Z; = 2H emzo COS (kt - mxo).
differential equation

dx dz In order to eliminate t, square the equations for (x - x;)

u(x,z,t 0 ) w(x,z,t 0 ) and (z - z;) and add the results. This gives:

Thus,
dx dz
k(Hj2)emz cos (kt 0 - mx) -k(H/2)emz sin (kt 0 - mx)
This is the equation of a circle of radius (Hj2)emzo. It is seen
or that the paths are circular and the radius tends to zero as
z 0 --+- oo. It will be seen in the linear wave theory that, at a
dz =-tan (kt 0 - mx) dx first approximation, one has X; ~ x 0 , z; ~ z 0 (Section 16-1),
and x 0 , z 0 can be considered the location of the fluid particle
If t 0 is taken as 0, this equation becomes: at rest.

dz =-tan ( -mx) dx =tan mx dx

1-4 Basic Equations
The integration of this equation gives
1-4.1 The Unknowns in Fluid Mechanics
emz cos mx = constant Problems
By varying the value of the constant the streamlines form In general, the density of a liquid is assumed constant so
the general pattern illustrated in Fig. 1-7. that equations are needed only for velocity and pressure.
The paths (or particle orbits) are defined by the differen- Hence, in the Eulerian system of coordinates, the motions
tial equation: are completely known at a given point x, y, z if one is able
to express V and p as functions of space and time: V =
~ = ____!!!____ = dt F(x,y,z,t) and p = F 1(x,y,z,t). Therefore, to solve problems
u(x,z,t) w(x,z,t) in hydrodynamics two equations are necessary, one of them
 Part 1: Establishing the Basic Equations that Govern Flow Motion
ro being vectorial. If V is expressed by its components u, v,
and w, four scalar or ordinary equations are necessary.
In free surface flow problems, the free surface elevation
11(x,y,z,t) around the still water level, or the water depth
h(x,y,z,t), is unknown and a kinematic condition is also
required. However, in that case the pressure p is known and
in general is equal to the atmospheric pressure.
For gases, two more unknowns need to be considered,
namely, the density p and the absolute temperature T.
Hence, to solve problems in the most general cases of fluid
mechanics, four equations are necessary. If V is expressed
by u, v, and w, then six ordinary equations are needed.
In hydrodynamics, basic equations are given by the
physical principles of continuity and conservation of
momentum. The equation of state and the principle of the
conservation of energy must be added in the case of com-
pressible fluid.
The reduction of a problem to such a small number of
variables (2 in hydrodynamics and 4 in gas dynamics), does
not occur for trivial reasons, but as a result of several
important arguments and assumptions. A number of
phenomenological functions are assumed to be known. For
example, it is assumed that the fluid is Newtonian and
either perfect or viscous, which defines the stress tensor.
The fluid obeys Fourier's law of conduction. Also, a
number of coefficients, such as heat conductivity, specific
heat, and viscosity, are supposed to be known functions
of the other unknown variables, such as density and/or
temperature.
1-4.2 Principle of Continuity
The continuity principle expresses the conservation of
matter, i.e., fluid matter in a given space cannot be created
or destroyed. In the case of an incompressible homogeneous
fluid, the principle of continuity is expressed by the conserva-
tion of volume, except in the special case of cavitation where
partial voids appear.
The continuity principle gives a relationship between
the velocity V, the density p, and the space coordinates
and time. If p is constant (in the case of an incompressible
fluid), it gives a relationship between the components of V
and the space coordinates, which are x, y, z. The equation
of continuity then becomes
au+ av + aw = 0
ax ay az
as it is demonstrated in Section 3-2.
It will be seen that V may be found in some cases of flow
under pressure, independent of the absolute value for p, from
the principle of continuity alone, but p will always be a
function of V except at the free surface.
1-4.3 The Momentum Principle
The momentum principle expresses the relationship
between the applied forces F on a unit volume of matter of
density p and the inertia forces d(pV)/dt of this unit volume
of matter in motion. The inertia forces are due to the natural
tendency of bodies to resist any change in their motion.
It is Newton's first law that "every body continues in its
state of rest or uniform motion via a straight line unless
it is compelled by an external force to change that state."
The well-known Newtonian relationship is derived from
his second law: "The rate of change of momentum is
proportional to the applied force and takes place in the
direction in which the force acts." F = d(mV)/dt.
In fluid mechanics this equation takes particular forms
which take into account the fact that the fluid particle
may be deformed. These equations will be studied in detail.
For an incompressible fluid, the integration of the momen-
tum equation with respect to distance gives an equality
of work and energy, expressing a form of the conservation
of energy principle.
If Vis expressed by u, v, w, then Newton's second law
has to be expressed along the three coordinate axes. This
gives the three equations
where pis assumed constant and Fx, FY, Fz are the com-
ponents of F along the three coordinate axes, respectively.

1-4.4 Equation of State
When considering a compressible fluid, two other
equations are required in addition to the equations ex-
pressing the two above principles. These are the equation
of state and the equation expressing the conservation of
energy.
The equation of state expresses the relationship which
always exists between pressure p, density p, and absolute
temperature T. For a perfect gas, this equation has the
very simple form
_P_=l _P_= 1
or
pgRT wRT
where R is the universal gas constant (R = 53.3 fttR for
air*) and w is the specific weight.
In a more general case of a real gas, it may take the form
p/pgRT = 1 + f 1 (T)p + f 2 (T)p 2 + where f 1 and f 2
are functions of the absolute temperature T only. In the
case of an incompressible fluid, the equation of state is
simply p = constant. The temperature can then be treated
as an independent variable having an (experimentally)
known significant influence on the coefficient of viscosity
only.
1-4.5 Principle of Conservation of
Energy
The next equation expresses the conservation of the total
energy (internal energy and mechanical energy). It is the
first law of thermodynamics.
The following equation is derived from this law in the
particular case of an adiabatic flow-that is, where no
heat is added or removed from the fluid mass. In that
case: p/l = constant, where k is the adiabatic constant
defined as the ratio of the specific heat at constant pressure
CP to the specific heat at constant volume Cv.
*oR are degrees Rankine, equal to degrees Fahrenheit plus 459.58.
Chapter 1 : Basic Concepts and Principles
In the case of isothermal flow at constant temperature II
which may necessitate the removal or addition of heat
from/to the fluid mass, p/ p = constant.
Inasmuch as hydrodynamic problems alone are being
considered in this book, it is not necessary to further
consider the equation of state and the equation of conserva-
tion of total energy. The density p will be assumed known
and constant and the temperature T will be a variable
without influence upon the phenomenon under considera-
tion. However, it is evident that the dissipation of energy
by viscous forces may create a (small) elevation of tempera-
ture which in turn modifies the characteristics of the fluid.
In general, these effects are of secondary importance in
hydrodynamics, and in particular, the coefficient of viscosity
f1 is considered as a known constant.
1-4.6 Boundary Conditions
1-4.6.1 It is evident that a general solution of the system
of equations described above does not exist, but in many
particular cases solutions can be found when the boundary
conditions are specified. There are three main kinds of
boundary conditions:
1. At a free surface where the pressure is known and
generally equal to atmospheric pressure. The cases of
wind-water wave interaction and impulses on the
free surface are special cases where the variation of
the pressure at free surface is taken into account.
2. At a solid boundary, since the fluid cannot pass
through or escape from the boundary.
3. At infinity when the motion tends to a known value.
In such a case, the known conditions at infinity are
considered as "boundary" conditions.
1-4.6.2 At the free surface the pressure is known, but
the location of this free surface with respect to horizontal
datum level is unknown in general. So two conditions must
be specified: a dynamic condition, stating the value of
pressure, and a kinematic condition, stating that the particle
at the free surface remains at the free surface.
 Part 1 : Establishing the Basic Equations that Govern Flow Motion

12 Since p is normally a constant at any time, the total z

differential of p(x,y,z,t) is zero; that is,

op op op op
dp = - dx +- dy +- dz +- dt = 0
ox dy az at

The total derivative of p with respect to t is given as:

dp = ap + op dx + ap dy + op dz = 0
dt ot ax dt oy dt oz dt

If the variables u = dxjdt, v = dyjdt, and w = dz/dt,

are used (see Section 1-2.2), the free surface dynamic
condition becomes in the most general case:
y

op op ap op
-+u-+v-+w-=0
at ox oy az Figure 1-9 Uniform flow in a rectangular channel.

This condition, involving a force, has to be used with

the equation expressing the momentum principle.
The kinematic condition will be developed in Section
16-1.3.2. For the time being, it is sufficient to know that if
z = 17(x,y,t)
is the equation of the free surface, the kinematic condition is:
For instance, the boundary conditions in the case shown
in Fig. 1-9 are:
u = 0 for x = 0 and ~ = x 1
w = 0 for z = 0
p = constant for z = z 1
More generally, if F(x,y,z) = constant is the equation of
the boundary, the following boundary condition expresses
the fact that the surface F and V are tangential at any point,

1-4.6.3 At fixed solid boundaries, friction reduces the oF oF aF

velocity to zero, so that V = 0. This condition is used
in the continuity equation, and since a friction force is
involved, it must also be used in the momentum equation.
If the fluid is assumed to be perfect (or ideal), only the
component perpendicular to the boundary is zero. The
component of the velocity V which is tangential to the
boundary remains. It is used primarily with the continuity
relationship. It does not involve a force but a continuity
statement: the fluid cannot pass through or escape from
the boundary (unless there is cavitation).
u-+v-+w-=0
ax ay oz
i.e.,
V grad F = 0
1-4.6.4 At movable solid boundaries (wheel of turbine,
wave paddle, etc.) the boundary condition expresses the
fact that the fluid follows the boundary (see Fig. 1-10).
Thus, the velocity component of the fluid perpendicular
to the boundary is equal to the corresponding component
 Chapter 1: Basic Concepts and Principles

_1 2A
,_ z 13
I I I
I I

Figure 1-10 A piston wave paddle gives a movable boundary Figure 1-12 Periodic gravity wave in infinite
condition.
depth: V--> 0 when z--> - oo.

of the boundary itself. The other component follows the
corresponding boundary motion component for a real
fluid only.
If F(x,y,z,t) = constant is the equation of the movable
boundary, the following boundary condition expresses
the fact that the fluid remains at the boundary:
aF aF aF aF
-+u-+v-+w-=0
at ax ay az
1-4.6.5 An infinite distance can give a boundary condition
if the motion tends to a well known value far from the
studied space. For example, consider the diagram shown
in Fig. 1-11. The motion is known at infinity and can be
written (as far as friction effect is negligible) V = constant
for x-+ oo.
As another example, it is well known that the motion
of swell in deep water is limited to a zone near the free
surface. Hence, the periodic gravity wave theory in infinite
depth is based on the boundary condition V -+ 0 when the
distance from the free surface tends to infinity: z -+ - oo
(Fig. 1-12).
PROBLEMS
1.1 Consider a two-dimensional flow motion defined by
the velocity components:
u =A+ Bt v= C
where A, B, and C are constant parameters. Demonstrate
that the streamlines are straight lines and that the particle
paths are parabolas.
1.2 A disk of radius R rolls without slipping on a horizontal
plane at a constant angular velocity k. Demonstrate that
the "streamlines" are circular and that the paths are
trochoidal.

Figure 1-11 Flow in a pipe past a diaphragm:
V = V0 when x --> oo.
1.3 Consider a fixed cylinder in a uniform current of
constant velocity. It will be assumed that there is no separa-
tion. Sketch the streamlines, the paths, and the streaklines
intuitively. Consider now a cylinder moving at constant
velocity in still water, and sketch the streamlines, the paths,
and the streaklines. Explain the differences between the
two cases, considered as a steady and an unsteady motion,
respectively.
 Part 1 : Establishing the Basic Equations that Govern Flow Motion

I4 1.4 Devise a general graphical method of construction to period; and d is the water depth.) Find the equations for
determine a steady flow pattern around a body moving the streamlines, and sketch them. Assuming Ix - x 0 I
at constant speed in a still fluid from the pattern of stream- and lz- z0 1 are small, find the approximate equation
lines defined with respect to a fixed coordinates system for the particle paths. Sketch these paths.
and vice versa.
1.6 Express mathematically the boundary conditions for
1.5 A two-dimensional flow motion (linear periodic gravity
any kind of flow motion taking place between the bound-
waves in water depth, d) is defined in a Lagrangian system
aries defined by Fig. 1-13. A hinged paddle will be assumed
of coordinates by the equations:
to have a small sinusoidal motion of amplitude e at the
_ . H cosh m(d + z) . (k _ ) free surface.
x - x, + 2 . h d Sill t mx
sill m
H sinh m(d + z) 1.7 Consider a two-dimensional body moving at velocity
z = z; + -2 . h d cos (kt - mx) U in the negative X direction. The nose of this body can
Sill m
be defined by a curve such that y = x 1 i 3 , and u and v
where H is the wave height; m, k, and d are constants. are the components of velocity along the body. Establish
(m = 2njL; L is the wavelength; k = 2n/T, Tis the wave the relationship between u, v, U, and y.

1##-fr#$#/ll.f$/.f;!l.f$,
-- - 1/11/l/l//l/l/lll/(1111}
b c
.-
0
------

Figure 1-13
Wave basin, top and side view.

~ ~REE SURFA~
 Chapter 1 : Basic Concepts and Principles

Then consider the case where the body is fixed and the 15
fluid is moving at a velocity U.

1.8 Consider a translatory wave in a channel moving

without deformation at a constant velocity C in the negative
X direction. At a given time t the wave profile is defined
approximately by the relationship z = Ax 112 where A is a
constant. Demonstrate that the free surface velocity
components us and ws are related by the equation
A2
Ws = (us - C) 22

1.9 A sphere of radius R is moving at a velocity U(u.,v.,ws)

through a fluid at rest. Establish the equation for the
boundary condition in the case of a perfect fluid.

1.10 Draw the streamlines and paths of monochromatic

periodic waves as given in Problem 1.5, relative to a system
of Cartesian coordinates moving at speed C = k/m in the
wave direction. The free surface is defined by z ~ 0 and the
bottom by z = -d. Are they streamlines?
16

Chapter 2 2-1 Introduction to the Different

Kinds of Motion
In mathematical terms, the motion of the fluid elements
Motions of along their own paths is considered as the superposition of
different kinds of primary motions. The physical interpreta-
tion of these motions is given first by considering the simple
Fluid Elements; case of a two-dimensional fluid element, where all velocities
are parallel to the OX axis and depend only on y (like a
Rotational and laminar flow between two parallel planes).
Consider the infinitesimal square element ABCD of area
lrrotational Flow dx dy at time t and the same element at time t + dt:
A 1 B 1 C 1D 1 (Fig. 2-1).
The velocity of A and D is u, and the velocity of B and C
is u + du = u + (oujoy) dy since AB = dy, and u in this
case is a function of y only.
It is possible to go from ABCD to A 1 B 1 C 1 D 1 in three
successive steps
1. A translatory motion which gives A 1 B 2 C 2 D 1 ; the
speed of translation is u
2. A rotational motion which turns the diagonals
A 1 C 2 and D 1B 2 to A 1 C 3 and D 1 B 3 , respectively
3. A deformation which displaces C 3 to C 1 and B 3 to B 1
If in the limit dt tends to zero, C 1 C 2 tends to zero. If this
occurs the angle C 2 C 1 C 3 will tend to 45 when dx = dy.
Hence

c c _ C1 Cz _ (oufoy) dy dt
2 3- J2 - J2
The rate of angular rotation is:

dr -!:__{segment} "'!:__ c2 c3 - !:__ c2 c3

dt - dt radius - dt A 1 C 2 - dt J'2dy

Introducing the value C 2 C 3 previously given, it is found

that the rate of angular rotation is:

dr 1 au
dt 2 ay
 Chapter 2: Motions of Fluid Elements: Rotational and Irrotational Flow

u + ~ dy \ B,
~y /-

r--
I
8,...-
________,c ____ -~{--\..:,;:8.:...'____,_;;,.-.,.;:-...,c,
I'
: ',
Figure 2-1

L
I
dy I
Elementary analysis of different kinds I
of motion of a fluid particle. I /
I,/
Ai, u dx _JD-----A, ~----~

Similarly, the rate of deformation would be found to be velocity components. The translation can be along a straight
equal to: line or a curved line.

~ (C3C1) _~au
at A1C3 2ay
If x, y, and z are the coordinates of A at time t, then
x + L\x, y + Lly, and z + Llz are the coordinates at time
t + M. The translatory motion is defined by the equations
In the general case, there are three major constituents of Llx = u Llt dx = u dt
particle motions and deformations. They are: Lly = v Llt or dy == v dt
1. The velocity components V(u,v,w): translation Llz = w Llt dz = w dt
2. The variation of velocity components in their own The flow of particles along parallel and straight stream-
direction: dilatation lines with a constant velocity (so-called uniform flow) is a
3. The variation of velocity components with respect to case of translatory motion only (Fig. 2-3).
a direction normal to their own direction: rotation The translatory motion may be defined more rigorously
and angular deformation as the motion of the center of the rectangular element
These three constituents are successively discussed in the
following sections. Figure 2-2 Translatory motion.
z
2-2 Translatory Motion
Consider the particle at the point A(x,y,z) at time t.
The point is a corner of a small rectangular element, the
edges of which are parallel to the three axes OX, OY, OZ,
respectively (Fig. 2-2). When the particle moves so that the
edges of the rectangular elements remain parallel to these
axes, and maintain a constant length, it is a translatory
motion only. This implies no space dependence of the y
 Part I : Establishing the Basic Equations that Govern Flow Motion

I8 and thinner. It is a case of dilatational or linear deformation

I+ dt
Figure 2-3 An example of translatory motion:
uniform flow.
instead of the motion of the corner of the element. However,
this change complicates slightly the development of figures
and equations and gives, finally, the same result. Hence in
the following discussion, translatory motion will be defined
as the motion of a corner.
In the following, the physical meanings and the cor-
responding mathematical expressions are studied in the
case of a two-dimensional motion at first, then they are
generalized for a three-dimensional motion.
2-3 Deformation
It is easier to explain this kind of motion with the aid of
an example. Two kinds of deformation have to be distin-
guished: dilatational deformation and angular deformation.
2-3.1 Dilatational or Linear
Deformation
In a converging flow, the velocity has a tendency to
increase along the paths of particles. Therefore, the velocities
of the edges perpendicular to vector V (or to the streamlines)
are not the same (Fig. 2-4). The particle becomes longer
superimposed on a translation only provided the angles
between the edges do not change.
Now consider the two-dimensional particle ABCD of
which the velocity in the x direction of the edge AB is u,
and the velocity of the edge CD is u + du = u + (oujox) dx,
since AD = dx (Fig. 2-5). Similarly, the velocity of AD in
they direction is v, and the velocity of BC is: v + (ovjoy) dy.
Note that here the derivatives of u with y or v with x are
not considered and the derivatives of velocity (oujox) dx
and (ovjoy) dy, do not depend upon time. The velocities of
dilatational deformations are (oujox) dx and (ovjoy) dy.
After a time dt, BC becomes B'C', the length BB' being equal
to the product of the change of velocity and the time, that is,
BB' = (ovjoy) dy dt. [The velocity (ovjoy) dy is negative in
the case of Fig. 2-5.] CD becomes C'D' and similarly DD' is
equal to: DD' = (oujox) dx dt.
The velocities of dilatational deformation are per unit
length:
(oujox) dx ou (ov/oy) dy ov
dx ox dy oy
The sum oujox + ovj oy is the tot~l rate of dilatational
deformation, i.e., the rate of change of area per unit area.
Areas BCEB' and D'C'ED must be equal in the case of an
Figure 2-5 Components of dilatational deformation.
bv
v + ~ dy
y
c
---r
B

Figure 2-4 Dilatational deformation offluid

particle in a convergent. oy Y
bv d dt

a' --------- E_- j __ _ _ __ , c'

y

L.
I
v bu I
u ut~ d~ I
I
I
'
L :~dxdl
~---'----~---------_J
A

_JD
 Chapter 2: Motions of Fluid Elements: Rotational and Irrotational Flow
Figure 2-6 Shear deformation in a bend.
incompressible fluid. Their difference gives the rate of
expansion or compression in the case of a compressible
fluid.
2-3.2 Angular Deformation or
Shear Strain
Angular deformation may be illustrated by the behavior
of a fluid particle flowing without friction around a bend.
It is a matter of common observation that it is windier
at a corner than it is in the middle of a street. In the similar
case of fluid flow around a bend, neglecting the effects of
friction, the velocity has a tendency to be greater on the
inside than it is on the outside of the bend. The law
:l ; '" r
V x R = constant may be approximately applied where
Figure 2-7
Angular or shear deformation.
0~----~- x
V is the velocity and R is the radius of curvature of the 19
paths. Hence, if the particle A is the corner of the rectangle
ABCD, the edge AB of the rectangle moves at a greater
velocity than the edge CD and there is angular deformation
(Fig. 2-6). This angular deformation is proportional to the
difference of velocity between AB and CD.
Now considering, for example, the case presented in
Fig. 2-7, in which the velocity of AB is u, and the velocity of
CD is u + du = u + (8uj8y) dy, then the distance CC'
(or DD') after a time dt is (8uj8y) dy dt. The angular velocity
IS
(8uj8y) dy au
dy 8y
Note that in contrast to the case of dilatational deformation,
the derivatives of u withy and v with x are retained here. The
derivative of the velocity (8uj8y) dy does not depend upon
time.
Similarly BB' (or DD") is equal to (8vjox) dx dt. When
these two deformations exist at the same time, the sum of the
angular velocities (oujoy) + (ovjox) is the rate of angular
deformation.
Note that 8ujoy was chosen equal to ovjox in Figure 2-7,
and the bisectors of the angles made by the edges of the
square element tend to remain parallel to their initial
bU
u+;-dy
0 y I I
- ' _ __,o_ - --7 o'
c r ---Tc_
I
I I
y ~dt I
by I I
I I
I
---
I
A L - - -- - -..J B
 Part 1 : Establishing the Basic Equations that Govern Flow Motion

20 positions during the angular deformation. When the
bisectors do not remain parallel to their initial positions,
the motion is said to be rotational.
2-4 Rotation
Although flow motions can be classified in various ways
according to some of their typical characteristics (such as
laminar or turbulent, frictionless or viscous, with or without
friction, steady or unsteady), one of the most important
divisions in hydrodynamics consists of considering whether
a flow is rotational or irrotational. Hence, the abstract
concept of irrotationality is fully developed in the following
sections.
2-4.1 Mathematical Definitions
For a two-dimensional motion, it has been shown that
the angular velocities of deformation are oujoy and ovjox.
The rotation of a particle is proportional to the difference
between these components. Indeed, if oujoy = ovjox,
there is angular deformation without rotation and the
bisectors do not rotate (Fig. 2-7). But if oujoy # ovjox, the
bisectors change their direction, and there is either both
rotation and angular deformation, or rotation only (Fig.
2-8).
The difference (oujoy) - (ovjox) defines the rate of
rotation, and therefore, a two-dimensional irrotational
motion is defined mathematically by (oujoy) - (ovjox) = 0.
Angular deformation can be considered without rotation
when (oujoy) - (ovjox) = 0 and (oujoy) + (ovjox) # 0, and
theoretically, rotation can exist without deformation when
(oujoy) - (ovjox) # 0 and (oujoy) + (ovjox) = 0. This case
is rare in practice, since rotation generally occurs with
angular deformation in physical situations. A forced vortex,
such as that schematically shown on Fig. 2-9, is a rare case
in which particles rotate without deformation. However,
this can be considered more as a special case of hydrostatics
where the centrifugal force is added to gravity, rather than
a real rotational flow.
2-4.2 Velocity Potential Function:
Definition
The concept of irrotational motion is very important in
hydrodynamics since many real flows are nearly irrotational.

I
I
'jl

Figure 2-8
Rotation and deformation.
SHEAR
DEFORMATION ROTATION ROTATION AND
WITHOUT ROTATION WITHOUT DEFORMATION DEFORMATION

6u 6v 6u
- ov ou - OV
ry - 6X = 0 dy 6X
;. 0 6y 6X
;. 0

6u 6v ;. 6u 6v 6u 6v ;.
6-y +6- 6-y + 6X YY + IX
0 0 0
X
 Chapter 2: Motions of Fluid Elements: Rotational and Irrotational Flow

where i and j are the unit vectors along the X and Y axis, 21
respectively. The magnitude of the velocity becomes

2-4.3 Theoretical Remark on

Irrotational Flow
It is useful to study the characteristics of an irrotational
flow. For this purpose, the previous example of a flow
without friction in a bend, or of a free vortex motion
defined by the equation VR = K, is analyzed (see Fig. 2-10).
Consider the rectangular fluid element ABCD between
two path lines defined by their radius of curvature R 1 and
Figure 2-9 Forced vortex (V = KR), rotation
without deformation. R 2 such that R 2 = R 1 + dR, dR being infinitely small.
After an interval of time dt, ABCD becomes A'B'C'D',
and
The properties of irrotational motion lead to a number of
simple and powerful analytical, graphical, or analog AA ' = CC' = V1 d t = Kdt-
methods which can be used in the solution of hydraulic Rt
problems. Most of these methods result from the existence BB' = DD' = V2 dt = K !!._
of a special function, the velocity potential. R2
The velocity potential is defined as a single-valued
function such that u = (8/8x) and v = (8j8y) [or Figure 2-10 The case of an infinitely small
alternately u = - (8/8x), v = - (8j 8y)]. If the functions displacement in an irrotational flow.
u and v are continuous, this function will satisfy the irrota-
tionality condition, which in two dimensions is (8uj 8y) o'
- (8vj8x) = 0. When the expressions for u and v are
substituted into the condition for irrotationality, the
results are

since the differentiation with respect to two variables is

independent of the order in which the differentiation is done.
The velocity potential will be shown to exist for three-
dimensional motion as well.
The value of the velocity V in terms of velocity potential
function is
 Part 1 : Establishing the Basic Equations that Govern Flow Motion
22 The side AB rotates to A' B' by an infinitely small range
r such that
BB' K dt
r :::::: tan r :::::: 0' B :::::: R 2 0' B
and
AA' K dt
r :::::: tan r :::::: - :::::: ~~-
O'A R 1 O'A
Equating these last expressions leads to:
R2 O'A
O'A- O'B = R2 - R1
R1 O'B
or O'B = R 1 and O'A = R 2 When these values are substi-
tuted into the equation for r the result is: r = (K dt/R 1 R 2 ).
Since dR is small, R 2 ~ R 1 and the equation can be written
as: r = (K dt/ R i). Since (} 1 is small, sin (} 1 :::::: (} t. and
8 1 = (AA'/R 1 ) = (K dt/Ri), hence r = 81 .
The side AC rotates into A'C' through the angle 8 1 .
Since the two sides AB and AC rotate by the same quantity
8 1, but in opposite directions, the bisector AX remains
parallel to the bisector A' X'. The orientation of this median
line remains unchanged, which is the condition for the
motion to be irrotational.
It must be emphasized that the previous demonstration
holds true only when an infinitely small displacement is
considered. It does not hold true for a finite displacement,
since the two bisectors have a tendency to rotate in the
same direction.
Both the angle of rotation of the bisectors and the angle of
angular deformation have finite values for a finite displace-
ment of the element. They both tend to be infinitesimal
when the displacement tends to zero. However, in an irrota-
tional motion, the angle of rotation is an infinitesimal of
higher order than the angle of deformation. In a real flow,
the irrotationality of the motion cannot be determined by
observing the deformation of a "particle" in motion along
its path since this characteristic is essentially local.
2-5 Practical Limit of Validity of
Irrotationality
2-5.1 Rotation Caused by Friction: The
Kelvin Theorem
2-5.1.1 In practice it is very important to know when
the motion of the fluid particles can be considered as an
irrotational motion. Only if the assumption of irrotation-
ality is valid can the powerful methods of calculation based
on velocity potential, conformal mapping, relaxation meth-
ods, flow nets, electrical analogy, etc, be applied successfully.
The concept of irrotationality is essentially mathe-
matical [(8u/8y) - (8vj8x) = 0 in the case of two-dimen-
sional motion]. The difficulty arises when one tries to
establish some simple practical rules for assessing the
validity of this assumption. Indeed, rotation is often caused
by viscous forces, but a rotational solution also exists for a
perfect fluid, and irrotational flows exist in a viscous fluid.
For example, let us consider a reservoir, where the flow
velocity is practically zero, and a connected duct. Initially
the fluid is irrotational, but viscous stresses eventually
cause the flow to become rotational at the entrance of the
duct; here friction forces cause rotation. This experimental
fact is translated mathematically by the Kelvin theorem
which applies in the case of a viscous fluid of constant
density, under a constant gravity force. An exact demon-
stration of the theorem is beyond the scope of this book
but a physical introduction to rotation will suffice in the
following.
2-5.1.2 It is easy to see whether a motion is physically
rotational or irrotational by a consideration of friction
effects. A physical understanding can be gained by the
following considerations. Near a fixed boundary, along
which particle velocities are zero, particles on adjacent
paths have significantly different velocities; hence, a line
joining at one time two particles on the same path will
rotate much less than that of a line joining two particles on
adjacent paths. The difference of direction of the friction
forces acting on the two opposite sides of the particle,
 Chapter 2: Motions of Fluid Elements: Rotational and Irrotational Flow
Figure 2-11 The variation of velocity in a direction
perpendicular to the flow induces a difference
in the direction of the friction forces
and a torque resulting in a r otational motion.
causes a torque resulting in a net rotation (Fig. 2-11).
Figures 2-12 and 2-13 illustrate some cases where it is
possible to see easily whether the assumption ofirrotational
motion is permissible.
In general, the motion can be assumed to be irrotational
when the velocity gradient is small (such as in periodic
gravity wave), when streamlines converge rapidly, and when
Figure 2-12
Examples of rotational and irrotational motion.
the velocity distribution depends on the shape of the 23
boundaries and not on their roughness. Motion is rotational
in the neighborhood of boundaries or in a diverging flow.
As mentioned previously, close to boundaries the large
velocity differentials between particles on adjacent paths
cause the motion to be rotational. A region of high velocity
gradients may be adjacent to a region where the velocity
gradients are sufficiently small for the motion to be treated
mathematically as irrotational. The region of high gradients
is called, for instance, a boundary layer if it occurs near
a solid boundary or between fluids of different nature
(liquid- gas interface) or shear layer between two fluid
streams. A motion may be considered irrotational only if
the boundary layer is of little importance, i.e., relatively
thin. Figure 2-14 illustrates the case of a weir where the
boundary layer thickness increases downstream. The
motion is irrotational only near the top.
2-5.2 Rotational Solutions in a Perfect Fluid
It is seen that the rotation may physically be due to the
friction. Physical considerations of friction effects have
resulted in practical rules to follow.
However, there exist mathematical solutions of rota-
tional motion where the friction forces are neglected. The
classical Bernoulli equation of elementary hydraulics is
valid only along a streamline when motion is rotational
without friction (see Chapter 10). One specific example of
nondissipative (i.e., without friction) rotational motion is
 Part I : Establishing the Basic Equations that Govern Flow Motion

24 IRROTATIONAL (Convergent)

I V ERY ROTAT IONAL

(Lines of Seporo llon)

VE NT

Figure 2-13
Examples of rotational and irrotational motion.

IRROTAT IONAL ZONES ( Stognont)

the Gerstner's theory on periodic gravity waves. In this
theory the paths of fluid particles describe circles. The
particles also rotate about themselves in the opposite
direction (Fig. 2-15). The results are expressed by an exact
mathematical solution of the basic equations in which the
friction terms have been neglected, but in which the inertial
Figure 2-14 Effect of the boundary layer.
rotational terms are taken into account exactly (see Section
17-1.4).
2-5.3 Irrotational Solutions in Viscous
Fluids
One also finds dissipative motions, which are considered
as irrotational. For example, friction forces have a pre-
dominant effect on such phenomena as the damping of a

Figure 2-15 Path and rotation of a fluid

particle in a Gerstner wave.

DIRECT ION OF
WAVE T RAVE L

- A I R ENT RAINMENT
 Chapter 2: Motions of Fluid Elements: Rotational and Irrotational Flow

gravity wave though a filter and the flow through a porous
medium. However, in these cases, only the mean velocity
with respect to space is considered. The actual system of
complicated rotational motions through the porous medium
is studied as an average motion with respect to space which
is irrotational at low Reynolds number (see Chapter 9).
Similarly, turbulent flow is strongly rotational but the
mean motion with respect to time may often be considered
as irrotational (see Chapter 8).
It may also happen that the flow is irrotational when the
sum of all the viscous terms which appear in the momentum
equation equal zero, although each term individually is
different from zero. Such kinds of motion are dissipative
and irrotational. A specific example of such a case is the
motion generated by a circular cylinder rotating steadily
about its axis in an unbounded viscous incompressible
fluid. The velocity gradient normal to streamlines can be 25
large near the cylinder. The motion is still irrotational.
The motion of a free vortex is the same whether one
considers the fluid perfect or viscous. The solution to the
momentum equation for a perfect fluid (VR = constant)
makes the sum of all viscous terms of the momentum
equation equal to zero.
2-5.4 Energy Dissipation, Shear
Deformation, and Rotationality
The fact that the motion is rotational does not necessarily
mean that it is dissipative. A motion is dissipative when
there are linear and/or angular deformations associated
with a non-negligible viscous coefficient. So an irrotational
free vortex can be dissipative, while a rotational forced
vortex is not dissipative.

o'

y+vdt

Figure 2-16
Two-dimensional coordinate system.
y + dy

C:''
v
X

y ----- t
A:-u 9
ou
u -t-dx
I ox
I I
I I
~--------~-----------J----------------J-------~- X
0 )( x + dx x+udt
 Part 1 : Establishing the Basic Equations that Govern Flow Motion

26
x + dx + u dt ou dx dt + -1 (ou
+ -ox - + -ov) dy dt - -1 (ov
- - -ou) dy dt
2 oy ox 2 ox oy
D' (2-2)
y + dy + v dt + ov dy dt + ~ (ou + ov) dx dt + ~ (ov - ou) dx dt
oy 2 oy ox 2 ox oy
'--,.--' "-.-' "-.-' '-.-' '-.-'
Initial Translation Dilatational Rate of angular Rate of
coordinates or linear deformation rotation
deformation
Angular or shear Rotation
deformation

Indeed, it will be seen in Section 5-5.3.2 that the viscous or

stresses are proportional to the coefficients of linear and
angular deformations presented in this chapter. Hence, the
viscous stresses are dependent on the existence of deforma-
tion and not rotationality.
2-6 Mathematical Expressions
Defining the Motion of a
Fluid Particle
2-6.1 Two-Dimensional Motion
Consider a fluid element ABCD at time t (Fig. 2-16).
The velocity components u and v are functions of x and y
such that du = (oufox) dx + (oufoy) dy and dv = (oufox) dx
+ (ovfoy) dy. At time t the space coordinates of A are x, y
and of D are x + dx, y + dy.
The coordinates of A and D at time t + dt are given in
Equation 2-1.
x + dx + u dt + (~~ dx + ~; dy) dt
{
D' y + dy + v dt + (~~ dx + ~>Y) dt
Adding and subtracting ~ovfox) dy dt to the x co-
ordinate and ~oufoy) dx dt to the y coordinate leads to
the form for the coordinates of D' expressed in Equation 2-2.
The physical meaning of the terms becomes apparent by
reference to the previous paragraphs.
2-6.2 Three-Dimensional Motion:
Definition of the Vorticity
Similar to the two-dimensional case, the coordinates of a
point D'(x + dx, y + dy, z + dz) of a three-dimensional
fluid element after a time dt become Equation 2-3.
x + dx + u dt + G~ dx + ~; dy + ~~ dz) dt

A'{xy ++ vudtdt y
ov ov ov
+ dy + v dt + ( dx dx + oy dy + jjz dz dt
) (2-3)
(2-1)
D'{x + dx + (u + du) dt ow ow ow )
y + dy + (v + dv) dt z + dz + w dt + ( ox dx + oy dy + oz dz dt
 Chapter 2: Motions of Fluid Elements: Rotational and Irrotational Flow

27
x + dx + u dt + -au dx dt + [1- (ov
ax
- + -au) dy + -1 (au
2 ax ay
- + -ow) dz + -1 (au
2 oz
- - -ow) dz
ax 2 oz ax
- -1 (av
- - -au) dy dt
2 ax oy
J
y + dy + v dt + -ov dy dt + [1- (ow
ay
- + -ov) dz + -1 (ov
2 oy az
- + -au) dx + -1 (av
2 ax
- - -au) dx
ay 2 ax ay
- -1 (ow
- - -ov) dz dt
2 ay az
J (2-4)

z + dz + w dt + -ow dz dt + [1- (au

az
- + -ow) dx + -1 (ow
2 az
- + -ov) dy + -1 (ow
ax 2 oy
- - -ov) dy
az 2 oy az
- -1 (au
- - -ow) dx dt
2 az ax
J

Adding and subtracting The coefficients of shear deformation will be defined as

1 ov 1 ow
and
--dydt
2 ox
--dz dt
2 ox f =~(ow+ ov) g =~(au+ ow)
2 oy az 2 oz ox
to the first line;
1 ow
- - dz dt
1 ou h= ~ (av +au)
and --dx dt 2 ax oy
2 ay 2 ay
to the second line; and The coefficients of rotation will be defined as
1 ou 1 ov
- -dx dt
2 oz
and - -dy dt
2 oz ~=~(ow_ ov) 1] = ~ (au - ow)
2 oy az 2 oz ox
to the third line leads to Equation 2-4.
The following notations will be used: The coefficients ( = ~ (av _au)
of dilatational deformation will be defined as 2 ax ay

ou b = ov ow The coordinates of the point D' are now written as Equation

a=- C=-
ox oy OZ 2-5, in which 2~, 2ry, and 2( are the components of a vector

x + dx + u dt + adx dt + (h dy + g dz) dt + (17 dz - ( dy) dt

y + dy + v dt + b dy dt + (f dz + h dx) dt + (( dx - ~ dz) dt (2-5)

z + dz + w dt + c dz dt + (g dx +f dy) dt + (~ dy - 17 dx) dt
'--.-' ~ ~
Initial Translation Dilatational Angular Rotation
coordinates deformation deformation
 Part 1 : Establishing the Basic Equations that Govern Flow Motion

28 which represents the vorticity of the fluid at a point. A The above equations would still hold, even if the velocity
three-dimensional irrotational motion is defined by ~ = 0, potential had been negative, so the velocity potential can
rJ = 0, and ( = 0; that is, also be defined by V = -grad .

aw av au aw av au 2-6.4 Stokes Analogy: Experience of

ay - az' az ax' ax ay Shaw
A three-dimensional rotational motion may be a two-
2-6.3 Velocity Potential Function in the dimensional irrotational motion when the rotation is always
Case of a Three-Dimensional Motion in the same plane. For example, a thin layer of water flowing
on a horizontal glass plate in which the thickness of the
The velocity potential function IS defined m three layer is very small in comparison with the other dimensions,
dimensions by has a rotational motion in vertical planes only (Fig. 2-17). If
a a a seen in the plane, the motion would appear as two-
u= - v=- W = - dimensional irrotational motion.
ax ay az
In the case of Fig. 2-17, motion in the vertical planes
This may be written vectorially as V = grad . XOZ and YOZ is rotational,~ and rJ =f 0, while the motion
When the values of the velocity potential are substituted in the horizontal plane X 0 Y is irrotational and ( =
into the equations for irrotational motion, the results are: J.(aujay - av;ax) = 0. It may be demonstrated that the
a2 0 2</J average velocity with respect to a vertical verifies similar
conditions of irrotationality.
ox oy = ay ox The streamlines seen in the plane are simply shown by the
This substantiates the definition of <P since <P always injection of dyes. The same result is obtained from a flow
satisfies the conditions for an irrotational flow. In other between two vertical parallel planes. This method is often
words, the existence of <P implies that the flow is irrotational. used to determine the flow pattern of two-dimensional or

THIN L AYER OF WAT ER

Figure 2-17
In a thin .flow of water, rotation exists
only in vertical planes.

GLASS PLATE
 Chapter 2: Motions of Fluid Elements: Rotational and Irrotational Flow

29

INJECTION Of' DYE

Figure 2-19

INTAKE
2.3 Determine the coefficients of dilatational and shear
deformation and rotation for a flow between two parallel
planes separated by a distance d = O.OlL. One of the
planes is assumed to be fixed, the other one moves at a
speed V = O.lLjt. The velocity distribution between the
INJECTION two planes is linear.
OF DYE

2.4 The velocity distribution of a laminar flow between two

parallel planes is given by the equation

u= __!__ a:yz - 2_ (a: - 2J1 v)Y

2J1 2J1 e2

Figure 2-18 Examples of studies based on Stokes analogy. where J1 is the coefficient of viscosity, e is the distance
between the two planes, a: is a constant equal to the head
loss or decrease of pressure per unit length: a: = dpjdx.

almost two-dimensional motion. Some examples are: flow

pattern around a wing, influence of an intake on the flow Figure 2-20
of a wide and shallow river (Fig. 2-18).

PROBLEMS

2.1 Consider a two-dimensional convergent section as

shown by Fig. 2-19. Determine the coefficient of linear
deformation at point x = 0, y = 0 where V = u = L/t
(Lis a unit length, tis a unit time).

2.2 Indicate the domains of Fig. 2-20, where the flow can be
considered as irrotational and the domains where the flow
is rotational. Give the reasons which prevail in your choice.
 Part 1 : Establishing the Basic Equations that Govern Flow Motion

30 V is the velocity of one of the planes, the other one being
assumed to be fixed.
Determine the coefficient of dilatational and shear
deformation and rotation as a function of y. Consider the
two cases where oc = 0 and V # 0 on one hand, and oc # 0
and V = 0, on the other hand, as two particular cases,
and explain their significance.
2.5 Express the velocity components as a function of a
potential function in a cylindrical (r,lJ,z) and spherical
(r,ci>,lJ) system of coordinates.
turbulent flow through rocks are
op = Ku2
ox
op = Kv2
ay
Demonstrate that such a flow is in general rotational.
2.9 Demonstrate that at a given location in a two-dimen-
sional flow, the value of angular rotation is independent of
the axis system of references, i.e.,

2.6 Derive the expression for irrotationality in polar system ! (av _ au) = ! (ov' _ ou')
of coordinates (r, lJ). The components of velocity are: 2 OX oy 2 ox' oy'
radial velocity v, tangential velocity v8 u' and v' being the velocity components along the x' axis
and the y' axis, respectively.
2.7 Consider two coaxial cylinders of radius R 1 and R 2
2.10 Demonstrate that
which are rotating at angular velocity w 1 and w 2 , respec-
tively. The velocity distribution of the fluid between these
two cylinders is given as a function of r by the expression
(R 1 < r < R 2 )

Determine the value of the coefficients of rotation and the

relationship between w 1 and w 2 which makes the flow
motion irrotational.

2.8 The equations for an average viscous flow through a

porous medium are defined by

op = Ku
ox
op = Kv
dy
where p is the hydrostatic pressure. Demonstrate that such
a flow is irrotational. The equations for an average fully
Chapter 3 3-1 Elementary Relationships
31

3-1.1 The Continuity in a Pipe

The principle of continuity expresses the conservation of
The Continuity mass in a given space occupied by a fluid. The simplest,
well-known form of the continuity relationship in elemen-
Principle tary fluid mechanics expresses that the discharge for
steady flow in a pipe is constant; that is, p VA = constant,
where A is the cross-sectional area of the pipe and V is the
mean velocity.
In the case of an incompressible fluid (p = constant) in a
uniform pipe (A = constant), the continuity relationship
becomes simply: V = constant. If the X axis is taken as
the axis ofthe pipe, then V = u, and the continuity principle
expressed in differential form becomes:

dV =au= O
dx ax
The case of a two-dimensional motion of an incompres-
sible fluid is now given. The general case will be treated
later.

3-1.2 Two-Dimensional Motion in an

Incompressible Fluid
Since no fluid is being added or subtracted during the
motion, the quantity of fluid involved is constant. This
may be expressed mathematically in the case of two-
dimensional incompressible motion. The development of
the mathematics follows.
Consider a rectangular element in two-dimensional
fluid motion as shown in Fig. 3-1. The rectangular bound-
aries have sides oflength a and band are considered to be fixed
with respect to the axes. It is not a moving fluid element.
The volume of fluid entering the left-hand boundary line
by unit of time is au 1 , and at the same instant, the amount
leaving the right-hand boundary line is au 2 The difference
in amount in the OX direction is thus: a(u 2 - ud =a !'!.u.
Similarly, the difference in amount in the 0 Y direction
is: b(v 2 - v1 ) = b !'!.v.
 Part 1 : Establishing the Basic Equations that Govern Flow Motion

y
32 After a time dt, because of the change of density with
respect to time, the quantity of fluid mass becomes

Hence the change of fluid mass in a time dt is

p dx dy dz-
ap dt ) dx dy dz = -at
( p +at op dt dx dy dz

X (3-1)
0 ~--------------------~

Figure 3-1 Rectangular element of an
incompressible fluid.
Since the total mass of fluid within the boundaries is
constant, the total loss must be zero: a du + b dv = 0;
that is, (du/b) + (dvja) = 0.
In the limit, when b and a approach zero, one obtains
(oujox) + (ovjoy) = 0. This differential form is permitted
because of the assumption of a continuous fluid. It should
be noted that oujox and ovjoy are the rates oflinear deforma-
tion of a fluid particle; hence, in an incompressible fluid,
the total sum of linear deformation is nil, as has been pre-
viously noted.
On the other hand, if one takes into account the change in
velocity and in density with respect to space coordinates, the
quantity of fluid mass entering through the section ABCD
during a time dt, parallel to the OX axis, is the product pu
times the area perpendicular to OX (ABCD) and the time dt.
Since ABCD = dy dz, the quantity of fluid mass entering is
pu dy dz dt. The derivative of u along AB and AD with
respect to dz and dy is of an infinitely small order and can be
neglected. Now the quantity of fluid mass coming out
Figure 3-2 Coordinate system for continuity
equation.
c G

3-2 The Continuity Relationship in the

General Case o bpu
: pu pu+bxdx
3-2.1 Establishment of the Continuity z ~

Relationship ~-------- F
I
I
Consider a fixed volume of fluid of which the edges dx, I
I
dy, dz are parallel to the axes OX, OY, OZ, respectively A
(Fig. 3-2). The continuity relationship is obtained by
considering that the change of fluid mass inside the volume X

dx dy dz during the time dt is equal to the difference between

the rates of influx into and efflux out of the considered
volume during the same interval of time.
The fluid mass at the timet is: p dx dy dz. y
 Chapter 3: The Continuity Principle

during the same interval of time through the section These continuity relationships can be written in a 33
EFGH is: shorter way as follows:

[ pu o(pu)
+ ---a;: dx
Jdy dz dt -op + d.IV pV
at
=
o
In the general case, both the density p and velocity u are or
assumed to be changed along dx. Hence the difference is
~ + p div V + V grad p = 0
[ pu
o(pu)
+ ---a;: dx
Jdy dz dt - pu dy dz dt =
o(pu)
---a;: dx dy dz dt
3-2.2 Physical Meaning and
Similarly, the difference due to the components of Approximations
motion parallel to the OY and OZ axes are, respectively, Consider, respectively, the three groups of terms:
o(pv) d d d d due to the difference of discharge
Tyxyzt op
across the sections BFGC and AEHD
at
(dx dz)

o(pw) d d d d due to the difference of discharge P(ou + ov + ow) or p div V

Tzxyzt across the sections AEFB and DHGC ax oy az
(dx dy) op op op
u-+v-+w- or V grad p
The total change of mass contained within the elemen- ox oy az
tary region during the time dt is
3-2.2.1 The first term, opjot, is the derivative of the
o(pu) o(pv) o(pw )] d d d d density with time at a given point. This term is nil in the
[ --+--+-- X y Z t (3-2)
ax ay az case of (1) incompressible fluid, since p is a constant; and (2)
a steady motion of a compressible fluid. This term has to
Equating Equations 3-1 and 3-2 yields be considered when sound, water hammer, shock waves,
etc. are studied.
- op dx dy dz dt = [o(pu) + o(pv) + o(pw)J dx dy dz dt
at ax oy az 3-2.2.2 The second group of terms is proportional to
the derivative of speed in the direction of motion at a given
Dividing both sides by dx dy dz dt yields
time. In the simple case of a three-dimensional motion of an
incompressible fluid

ou + ov +OW= 0 or div V = 0
ax oy az
Since [o(pu)]jox = p(oujox) + u(opjox), and similarly
for the terms o(pv)joy and o(pw)/oz, the continuity relation- When div V > 0, an expansion of the fluid is indicated,
ship becomes and conversely, div V < 0 signifies a compression.

op (au ov op op op
-+p -+-+- +u-+v-+w-=0
ow) 3-2.2.3 The third group of terms is proportional to the
at ax oy az ax oy az derivative of density with respect to the space coordinates
 Part 1 : Establishing the Basic Equations that Govern Flow Motion

34 at a given time. This derivative is usually negligible in
comparison with the others.
For example, consider a unidimensional sinusoidal
acoustic wave. The density is given as a function of time
and distance by the relationship p = A sin (Ct + x). The
derivative of p with respect to time t is opjot = AC cos
(Ct + x), and with respect to the distance x along the OX
axis, opjox = A cos (Ct + x). Hence,
This phenomenon occurs when a large vanatwn of
pressure results in the celerity of the wave exceeding the
velocity of sound. A shock wave travels at a higher speed
than usual pressure waves such as acoustic sound or water
hammer. Hence, when a supersonic flow or the effect of an
underwater explosion is studied, the derivative of the
density with respect to the distance has to be taken into
account in the continuity relationship.

opfox
3-2.2.4 Table 3-1 summarizes these considerations.
op/ot c
Unless otherwise specified, incompressible fluids only are
where C is the wave velocity. considered in the rest of this text.
Since Cis usually large compared to the particle velocity
u, u(opjox) is usually negligible by comparison with opjot.
However, there is an exception to this rule: the case of 3-3 Some Particular Cases of the
shock waves where the variation of p with respect to space Continuity Relationship
is theoretically infinite at the front of the wave.
The continuity relationship is often used in other forms
Table 3-1 in hydraulics. These forms are not so general but more
adapted to integration for the phenomena to be studied.
Some examples of the different forms used are provided
Uniform (one-dimensional) au= 0
flow of an incompres- ax by the case of unsteady flow, mainly unidimensional either
sible fluid at a free surface (channel, river) or under pressure (pipe,
gallery).
Two-dimensional flow of an
incompressible fluid
An unsteady free surface flow resulting in a change oflevel
with respect to time and space and caused by gravity
Three-dimensional flow of an au + av + aw = O action is called a gravity wave. Some examples are flood
incompressible fluid ax ay az waves in a river, bore (translatory waves), tsunami waves
due to earthquake, tides, harbor oscillation, and seiche.
div V = 0
An unsteady flow under pressure resulting in a change of
pressure with respect to time and space, caused by the
Unsteady motion in a ap + p(au + av + aw) = 0
compressible fluid at at ax ay az pressure gradient is called a pressure wave. Two examples
usual speed (acoustic are water hammer and acoustic wave. Such gravity waves
wave, water hammer)
~~ + p div V = 0 and pressure waves are studied from special forms of the
continuity relationships. They are valid when the distri-
Unsteady motion in a ap + a(pu) + a(pv) + a(pw) = 0 bution of velocity in a cross section can be assumed to be a
compressible fluid at at ax ay az constant; then the gravity waves are called long waves.
high speed (shock wave) Here, the continuity relationship is obtained by stating
~ + div (pV) = 0 that the change of volume of water during the interval of
time dt between two finite cross sections separated by the
 Chapter 3: The Continuity Principle

infinitely small distance dx is equal to the difference between Equating Equations 3-3 and 3-4 and dividing by dx dt, one 35
the influx and efflux from the considered volume during the obtains (ahjat) + (aqjax) = 0, (ahjat) + (ahujax) = 0, or
same interval of time
ah + u ah + h au = 0
at ax ax
3-3.1 Translatory Gravity Waves
Consider the volume defined by the two cross sections x 3-3.2 Irrotational Flow
and x + dx and the free surface at time t (Fig. 3-3). The If the density p is a constant, the continuity relationship
volume of influx, during a time dt, into the considered has been seen to be (aujax) + (avjay) + (awjaz) = 0.
volume at x is q dt or hu dt, where q is the discharge, h the In the case of irrotational motion, a velocity potential
depth, and u the horizontal velocity component. function has been defined by the relationships
The efflux out of the volume at x + dx is
a a a
[h J U=- V=- w=-
ax ay az
or u+~
a(hu) dx dt
Hence, introducing these expressions into the continuity
relationship yields
Hence the change of volume ~ V between these cross
sections x and x + dx is a difference az az az
ax 2 + ay 2 + az 2 = 0
aq ~V a(hu) dx dt (3-3)
~v = - -dxdt or = -
ax
which can be written '\1 2 = 0. This is the well-known
ax
Laplace equation which has been subjected to extensive
On the other hand, the volume at the time t is h dx, and research in mathematical physics.
at the time t + dt, since the free surface level changes the
3-3.3 Lagrangian System of Coordinates
volume is [h + (ahjat) dt] dx. Hence the change of volume
during time dt is: It has been explained in Chapter 1 that it is possible to
study problems in hydraulics either in Eulerian coordinates
ah or in Lagrangian coordinates (See Section 1-3.1). Since
~v = atdtdx (3-4)
this last system of coordinates is rarely used, the continuity

----t--
Figure 3-3 C:>h
1
h+bi dx I
Translatory wave.
I TRANSLATORY
I WAVE PROFILE
I
I
0 x tdx
 Part I : Establishing the Basic Equations that Govern Flow Motion

36 relationship is given here without any comment for the (1) is irrotational; (2) satisfies the continuity equation; (3) is
simple purpose of recognition in literature: such that a/az = 0 for z = -d.
a(x,y,z) =1 3.7 Derive the continuity equation and V2 for an incom-
a(x 0 ,y0 ,z0 ) pressible liquid in spherical polar coordinates {r,O,CI>) by
where x 0 , y 0 and z 0 are the coordinates of the considered considering a small volume bounded by the surface:
particle at time t 0 and x, y, z at time t. r, r + dr; 0, 0 +dO; Cl>, Cl> +del>.

3.8 Consider a uniform cylindrical pipe of diameter D, wall

PROBLEMS thickness e, and elasticity modulus E. The bulk modulus of
elasticity of the fluid is K. It will be assumed that under
3.1 Demonstrate that the continuity equation for a stream the influence of a variation of pressure apjax, the pipe is
tube can be written as: elastic and the fluid compressible. Demonstrate that the
continuity relationship is
a(pA) + a(pA V) = O
at as ap = (_!__ + _Q_)- 1 au
at K eE ax
where A is the cross section of the stream, and ds an element
of streamline. 3.9 Demonstrate that the continuity relationship for long
wave (tidal motion) in a spherical system of coordinates is
3.2 Consider the two-dimensional motions defined by
their velocity components u = A and v = 0 on one hand, a17 1 {a . a }
and u = Ax + B and v = 0 on the other hand, where A at + a sin Cl> aci> [h(O,CI>)U E sm Cl>] + ao [h(O,<I>)U sJ = 0
and B are different from zero. Calculate the divergence,
and tell in which case the fluid is compressible. where a is the mean earth radius, '7 the free surface elevation
around the still water level, UE is the component of the
3.3 Establish the continuity equation in a cylindrical system current velocity in the direction of increasing 0, us is the
of coordinates (r,O,z). A polar element of volume r dr dO dz component of velocity in the direction of increasing Cl>,
0 is the longitude, Cl> is the latitude, h(O,CI>) is the water depth.
will be used. The velocity components will be v, along a
radius, v6 perpendicular to v., and Vz along the axis OZ.

3.4 Express the Laplace equation V2 = 0 in a cylindrical

system of coordinates (r,O,z).

3.5 Establish the continuity equation for a stratified fluid of

density varying with z asp = k/z (ap;at will be assumed to
be zero).

3.6 Verify that the motion defined by the potential function

k cosh m(d + z)
<P = -a- h d cos (kt- mx)
m cos m
Chapter 4 4-1 Mass, Inertia, and Acceleration
37

4-1.1 The Newton Equation

To cause the motion of a constant mass M, or, more
Inertia Forces generally, to change the state of an existing motion, it is
necessary to apply to this mass a force F, which causes
an acceleration dV/dt such that F = M(dV/dt). This is a
vector relationship, i.e., true for both magnitude and
direction. The product M(dV/dt) is the inertia force, which
characterizes the natural resistance of matter to any change
in its state of motion.
The considered mass M is the mass of a unit of volume
of fluid
M = p (unit of volume) = p
where p is the density. Hence the fundamental equation
of momentum has the form F = p(dV/dt). Its three com-
ponents along the three coordinate axes OX, OY, OZ are
p(du/dt), p(dv/dt), and p(dwjdt), respectively.

4-1.2 Relationships between the

Elementary Motions of a Fluid Particle
and the Inertia Terms
To each kind of motion of the fluid particles (Chapter 2)
there corresponds an inertia force. The relationship
between the kind of motion described and the corresponding
inertia force is straightforward.
The elementary components of velocity of a fluid particle
as given in Chapter 2 are, in the case of a two-dimensional
motion,
Translation u, v

Dilatational deformation au dx -ov dy

ox ay
Shear deformation ! (au+ ov) dy
2 ay ax
1 (au av) dx
2 oy +OX
Part 1 : Establishing the Basic Equations that Govern Flow Motion
Rotation - ~ (av - au) dy
2 ax ay
~ (av _au) dx
2 ax ay
To each of these velocity components corresponds a
component of acceleration, which multiplied by p, yields
a component of inertia force.
Two types of inertia forces may be distinguished,
depending on the type of acceleration or elementary motion
considered. These are:
1. Local acceleration -corresponding to a variation of
the velocity of translation or the derivative of velocity
with respect to time.
2. Convective acceleration-corresponding to a varia-
tion of velocity of deformation and rotation or
derivative of velocity with respect to space.
The physical meaning of these accelerations and the
corresponding inertia forces is first examined; then their
mathematical expression is demonstrated. Chapter 5
deals with the applied forces F which have to be equated
to these inertia forces to obtain the momentum equation.
4-2 Local Acceleration
Local acceleration characterizes any unsteady motion,
i.e., motion where the velocity at a given point changes with
respect to time. Local acceleration results from a change
in the translatory motion of a fluid particle imposed by
external forces F.
4-2.1 Examples ofF/ow with Local
Inertia
Local acceleration occurs in the following cases.
4-2.1.1 In the first case, the velocity stays in the same
direction along a straight line and changes in magnitude.
If the velocity increases at a given point, which involves a
positive local acceleration, the inertia of the mass of fluid
in motion tends to slow it down.
-
WAVE TRAVEL
Figure 4-1 Periodic gravity wave in deep water.
The velocity at a given point changes in
direction only.
Alternatively, if the velocity decreases, which corresponds
to a negative local acceleration, the inertia of the mass
of fluid in motion tends to maintain the original velocity.
This type of phenomenon is observed in pipes or tunnels,
where a fluid stops or starts or balances because of a gate
movement. Local inertia has to be taken into account in
hydraulic engineering applications such as surge tanks,
water hammer, and locks.
4-2.1.2 In the second case, the velocity maintains the
same magnitude, but changes its direction. In this case
the inertia force is due to the centrifugal acceleration.
For example, in a periodic gravity wave in infinite depth,
the magnitude of the velocity at a given point is a constant
but its direction revolves continuously at all points (Fig. 4-1 ).
4-2.1.3 In the third case, the velocity changes at a given
point both in magnitude and direction. Some examples
of this case are turbulent flow (this important case is
fully developed in Chapters 7 and 8); alternate vortices;
displacement caused by a ship in motion; bore and tide
in an estuary; and periodic gravity waves in shallow water.
4-2.2 Mathematical Expression of Local
Inertia
The mathematical expression of the inertia forces
caused by a local acceleration is given by the change in the
velocity of the translatory motion with respect to time only.
The corresponding inertia force is equal to p(av;at) of
 Chapter 4: Inertia Forces

which the components along the three axes are, respectively: The expressions aujax, ovjoy, and owjaz are taken at a 39
p(oujot), p(ovjot), and p(owjot). The derivatives with respect given time, as seen in section 2-3.1. The corresponding
to space are not taken into account. acceleration is
_t1_ [au dx] = ou dx
dt ax ax dt
4-3 Convective Acceleration
Two similar expressions result for w and v. If u = dxjdt,
Convective acceleration characterizes any nonuniform v = dyjdt, and w = dzjdt, are substituted in these expres-
flow, i.e., when the velocity at a given time changes with sions and the result is multiplied by the density, the inertia
respect to distance. It is sometimes called field acceleration. forces are obtained. They are
Convective acceleration results from any linear or angular
au 1 a(u 2 )
deformation, or from a change in the rotation of fluid pu-=-p--
particles, imposed by external forces F. ax 2 ox
ov 1 a(v 2 )
4-3.1 The Case of Linear Deformation pv- = -p--
oy 2 ay
4-3.1.1 In a convergent pipe, it has been seen that the ow 1 o(w 2 )
velocity of a fluid particle, although constant with time at a pw-=-p--
az 2 oz
fixed location, tends to increase along the converging
streamlines. The velocity of the fluid particle increases It should be noticed that the last group of expressions
with respect to space. This is a positive convective accelera- may be written as (ajox)(pu 2 j2). This shows that the inertia
tion. The fluid tends to resist this acceleration by convective force is equal to the derivative of the kinetic energy with
inertia. respect to space along the three direction axes OX, OY,
In a divergent conduit, the velocity decreases and the and OZ, respectively.
fluid tends to continue its motion with the same velocity
4-3.2 The Case of Shear Deformation
because of its inertia. The applied forces cause a negative
convective acceleration. 4-3.2.1 In a bend, where the fluid particles are angularly
Expansion or contraction of a compressible fluid is the deformed, the fluid paths are curved and because of its
sum oflinear deformations and also results in corresponding inertia, the fluid tends to continue along a straight line.
inertia forces. This causes a centrifugal force proportional to the change
of direction which is imposed by the applied forces.
It is possible for the velocity of a fluid particle to keep
4-3.1.2 It has been seen that the linear deformation
the same magnitude along its path, but with a change in
velocity components are those given in Equation 4-1.
direction. This is the case of free vortex motion.

~~dx}
4-3.2.2 It has been seen that the velocity components
Two-dimensional of angular deformation for a two-dimensional motion are
motion
-av d y Three-dimensional
(4-1) ~(au+ av) dy
oy motion 2 ay ax

ow dz 1 (au av)
az 2 ay +ox dx
 Part 1 : Establishing the Basic Equations that Govern Flow Motion

40 Hence, as in the previous case, using the substitutions are the velocities of the components of rotation in a two-
u = dx/dt and v = dy/dt, the corresponding inertia forces dimensional motion, the corresponding inertia forces
become: obtained are

1
2 pv
( au av)
oy + ox
- ~ pv(av - au)
2 ax ay

~ pu(ov _ ou)
2 ax ay

4-3.3 The Case of a Change of Rotation
4-3.3.1 In the entrance to a pipe (Fig. 4-2), because of the
change in friction forces, there is a variation of rotation
of the fluid particles. Hence there are inertia forces cor-
responding to the natural resistance of the fluid to change
its rotational motion. In a uniform pipe, the rotation of
particles exists but there is no change in rotational magni-
tude and the corresponding acceleration is zero.
4-3.3.2 As in the two previous cases, since
- ~2 (av- au) dy
ax ay
~ ( av _au) dx
2 ax ay
It has been shown that it is possible to assume that the
motion is irrotational when friction effects are negligible.
It is evident that the same conditions lead to neglect of
rotational inertia forces.
4-4 General Mathematical Expressions
of Inertia Forces
4-4.1 Local and Convective Acceleration
In the general case both local acceleration and convective
acceleration occur at the same time. A simple example is
when a fluid oscillates in a nonuniform curved pipe. Hence,
in the general case, V and its components u, v, and w are
functions of both time and space coordinates. For example,
u(x,y,z,t). The total differential of u is

Figure 4-2 Zone of acceleration of rotation. au au au au

du = - dt + - dx + - dy + - dz
at ax ay az

The acceleration in the x direction is thus given by the total

differential of u, with respect to time:
~--~ du au au dx au dy au dz
IRROTATIONAL ROTAT IONAL
- = -at + -ax -+ - - +- -
MOT ION MOTION
dt dt ay dt az dt

Similar expressions occur for dv/dt and dw/dt.

Substituting u = dx/dt, v = dy/dt, and w = dz/dt, and
multiplying by the density p, the inertia forces given by
Equation 4-2 are obtained.
 Chapter 4: Inertia Forces

(auat +
ou ou ou)
u-+v-+w-
expression, valid along the 0 X axis: 4I
P ax oy oz ( ou ov ow)
[au
ot u-+v-+w-
p-+ ax ax ax
ov ov ov)
u-+v-+w-
Pev
ot + ax ay oz (4-2)
+ v(ou _ ov) + w(ou _ ow)]
oy ax az ax
(ow ow ow ow)
u-+v-+w- But
Par + ax oy oz aw = -1 -
au + vav- + w- a-
a (u 2 + v2 + w2 ) = - V2
u-
'--,-1
Local Convective
OX OX ox 2 ox ox 2
acceleration acceleration When the coefficients of the rotational vector
terms terms
2'1 =(au _ ow)
4-4.2 Elementary Acceleration oz ox
Components
2( = (ov _ ou)
Following a procedure similar to that used in the study ax oy
of the elementary motions of fluid particles (Section 2-6.2), are introduced, the following expression for the inertia
that is, adding and subtracting !pv(ovjox) and !pw(owjox) forces along the 0 X axis results
to the first line above, gives Equation 4-3, which emphasize
the previous physical considerations.
Similar forms can be obtained for they and z components
p[~~ + :X (~2 ) + 2(1]W - (v)J
of the forces. Similarly, it may be found that the inertia forces along the
OYand OZ axes are
4-4.3 Separation of Rotational Terms
It is often useful to transform the acceleration terms p - [ov a (V
at + -oy -2
2
) + 2((u - ~w) J
to a form which emphasizes both the kinetic energy
terms and the rotational terms. Adding and subtracting
p[v(ovjox) w(owjox)]
+ to the first line, gives the following
p - [ow + -a (V
at az 2
2
- ) + 2( ~v - IJU) . J

[au ou
p- + u- + (4-3)
at ox
'-v--1 '-v--1
Local Acceleration Acceleration Acceleration in
acceleration in linear in angular rotation
resulting in deformation deformation
a change in
translatory Convective acceleration terms
motion
Part 1 : Establishing the Basic Equations that Govern Flow Motion
These three expressions may be written more concisely
in vector notation as shown in Equation 4-4.
av
(
pat+ grad 2
V2
+ curl Vx V). (4-4)
~ '-.,-' '-.-'
Local Kinetic energy Rotational
acceleration term term
Convective acceleration
It has to be noticed that the convective inertia term
p grad ~z =p[i :x (~2) + j ;y (~2) + k ;z (~2)J
is, in fact, the derivative with respect to space of the kinetic
energy, pV 2 /2, of the particle.
4-5 On Some Approximations
4-5.1 Cases Where the Local
Acceleration Is Neglected
4-5.1.1 A rigorous steady motion never exists. There
is always a beginning and an end. However, many motions
in hydraulics are actually very close to being steady during
a given interval of time. In this case, since V does not vary
with time, the corresponding inertia term av jot is zero.
(The very important case of turbulent motion will be
studied in Chapters 7 and 8.)
However, there are many unsteady motions in hydraulic
engineering in which the local acceleration and the cor-
responding inertia terms are neglected.
This occurs when the velocities are slow and their
variations with time are very slow. For instance, in the
case of a periodic motion in which the period Tis very long:
oVjot ~ V/T Hence p(oVjot) would be negligible m
comparison with other forces.
Some particular cases where this approximation is
valid are (1) flow in a porous medium: variation of the
ground water table with respect to time; (2) flood wave
in a river; (3) variation oflevel in a reservoir because of the
variation of the upstream river flux, the spillway and bottom
outlet control, and turbined discharge; and (4) emptying
of a basin by a small valve. In all these cases, the flow is
considered as a succession of steady motions and calculated
as such without taking account of local inertia.
4-5.1.2 As an example of unsteady motion analyzed
as a succession of steady motions, the variation of level
in a basin is studied.
Consider the emptying of a rectangular basin of hori-
zontal cross-sectional area A. The volume of water above
a small hole of area S is A times z. The variation of z as a
function of time is given by the differential equation
Adz = Q dt where Q = CdSV, Cd being the coefficient
of discharge and V being given by the formula of Torricelli:
V = (2gz) 112 . The formula of Torricelli is well known in
elementary hydraulics, but it is important to point out
herewith that this formula is only valid when the local
inertia is negligible.
Introducing the value of Q = CdS(2gz) 1 12 in the above
equation and integrating gives the total time required to
empty such a basin, (z 0 is the initial depth)
- 2A 1/2
T- CdS(2g)ll2 (zo)
If S were large and A small, it would be necessary to take
account of the local inertia to calculate T.
Another similar example, previously cited, is that of the
variation of level in a reservoir of horizontal section S(z)
because of the variation of the upstream river flux. The
correspondi"ng calculation of the economical height of the
dam, the number of turbines and the spillway capacity
are deduced from Equation 4-5.
4-5.2 Cases Where the Convective
Acceleration Is Neglected or
Approximated
4-5.2.1 The local inertia term is proportional to the
velocity V and thus, it is a linear term. The convective
acceleration terms are quadratic and are proportional to
V 2 (or a product u2 , v2 , w2 , uv, uw, vw). Since the convective
 Chapter 4: Inertia Forces

43
Q(t) Llt = S(z) Llz + Q1 Llt + Q.(z) Llt + f(S) Llt (4-5)
'-v--' ~ '-v--' ~ ~
Upstream Change of Turbined Volume Loss by
influx of level in volume over the evaporation
the river reservoir spillway

acceleration introduces a quadratic term, the general 4-5.2.4 Sometimes only some terms of the convective
equation of momentum is nonlinear. acceleration may be neglected. The case of a two-dimen-
It is well known that it is easy to mathematically solve sional boundary layer on a flat plate is given here as an
many linear differential systems of equations. But it is example (Fig. 4-3). This example is particularly helpful
often difficult to solve a nonlinear system. This is the chief in understanding how the mathematical simplifications
cause of difficulty in fluid mechanics. For this reason, it is may be based on physical considerations. Hypothesis:
helpful to know when it is possible to neglect this quadratic u is large in comparison with v; the derivatives with respect
term. to y are large compared to the derivatives with respect
When V tends to zero, a quadratic term proportional to to x. The continuity equation for two-dimensional motion
V 2 tends to zero more rapidly than a linear term pro- shows also that oujox and ovjoy are of the same order.
portional to V. Hence, in practice, when V is small, V 2 is (Section 3-1.2). Hence, the 0 Y components of the convective
negligible and the convective inertia term is negligible in inertia terms are negligible because
comparison with the other terms expressing the local
inertia and applied forces. Such motions are called "slow
au au ov ov
u-+v-~u-+v-
motions." ox ay ax ay
Some examples of such motion are: (1) periodic gravity Similar approximations are made to analyze the develop-
wave theory (first order of approximation); (2) flow in a ment of a jet, and in the nonlinear long wave theory
porous medium, which obeys the linear law of Darcy (Chapter 18).
(such motion is defined only by an equality of applied forces
since the local acceleration is negligible); and (3) motion
of a small sphere in a viscous fluid (Stokes' formula). Figure 4-3 Introduction of simplifying
assumptions in the theory of
4-5.2.2 Sometimes a partial effect of the convective development of a boundary layer.
acceleration is taken into account by the use of an approxi-
mate solution given by a number of terms of a series.
(Example: gravity wave theories to the second order,
third order, etc. of approximation; laminar boundary layer v
theory; etc.)

4-5.2.3 Another method to take account of a partial

effect of the convective acceleration is by the assumption
of irrotationality when the friction effects are negligible
as seen previously.
 Part I : Establishing the Basic Equations that Govern Flow Motion

44 4-5.2.5 Linearizing the quadratic terms consists, for

instance, of substituting for the quadratic terms, y = AV 2
or dy = A 1 V dV, the linear terms, y 1 ~ BVor dy 2 ~ B 1 V
or dy 2 ~ B 2 dV, such that B = mean value of AV, B 1 =
mean value of A 1 dV, and B 2 =mean value of A 1 V.
This method is applicable when V or dV varies within a
small range; otherwise, the value of B, Bl> B 2 must be
changed and the problems solved step by step. Surge
tank stability calculations, laminar flow stability study ~nd
motion of a sphere in a viscous fluid (Oseen's theory) are
some examples where this method is used. The Oseen
theory is an attempt to improve the Stokian theory for the
flow around a sphere by taking into account some linearized
effects of convective inertia.

4-6 Geostrophic Acceleration

An additional inertial force due to the rotation of the
earth is now considered. This force is important in the study
of tidal motion, oceanic circulation, and storm surge.
(The storm surge phenomena is the piling up of seawater
along the coast due to the wind stress generated by storm
and hurricane.) It is called the Coriolis effect, which, in this
case, is due to the geostrophic acceleration.
4-6.1 The Coriolis Effect
4-6.1.1 Different parts of the earth rotate at different
speeds, depending upon their distance from the earth's axis.
Hence a fluid particle in the northern hemisphere moving
south, toward the equator, finds itself over a portion of the
earth which rotates faster than where it comes from. There-
fore, it appears for an observer fixed to the earth to fall
behind the rotating earth and the seabottom underneath.
Since the earth rotates from west to east, the deflection is
to the west, i.e., to the right of the direction of motion with
respect to the earth. The fluid particle will describe a path
westward (Fig. 4-4). Similarly, a fluid particle moving
northward from the equator has, by inertia, an excess of
eastward velocity relative to the seabottom underneath,
which is moving slower. Consequently, its path turns
EARTH'S
AXIS
Figure 4-4 The Coriolis effect in the northern
hemisphere.
eastward and the deflection is again to the right of the
direction of motion. In the sou~hern hemisphere, the same
reasoning applies, and the deflection is always found to the
left of the direction of motion (Fig. 4-5).
4-6.1.2 Let us now consider a fluid particle which rotates
with the earth. Its velocity about the earth's axis creates
a centrifugal force. Since the fluid particle moves at the
same speed as the earth, this force just balances the com-
ponent of gravity force which is perpendicular to the earth's
axis. If now it moves eastward with respect to the earth
along a parallel in the northern hemisphere, the particle
experiences an additional centrifugal force. Since the
particle is moving faster than the speed for which the
balance is possible, the particle tends to move further
away from the earth's axis. Therefore, it is deflected towards
the equator, where even though the distance from the
earth's center is the same, the distance to the earth's axis
is increased. This deviation tends to restore the balance

EARTH'S
AXIS
Figure 4-5 The Coriolis effect in the southern
hemisphere.
of forces (Fig. 4-6). For a fluid particle moving westward
the centrifugal force is not sufficient to balance the com-
ponent of gravity force. Therefore, the fluid particle is
deflected toward the Pole.
In the northern hemisphere, the deflection is to the right,
and in the southern hemisphere, the deflection is to the left.
4-6.1.3 In the case where a component of motion is
perpendicular to the earth's axis, a similar effect is observed.
Angular momentum is conserved. Therefore, if the particle
goes downward, towards the center of the earth, it will
be rotating faster than an earth fixed particle at the same
location and therefore it will deflect eastward. If the particle
goes upward, away from the center of the earth, it will
not be rotating sufficiently fast, and therefore will deflect
westward. Since most oceanic motions have very small
velocity components, this effect is most often neglected.
All these deviations are called the Coriolis effect.
Chapter 4: Inertia Forces
45
DIRECTION OF THE
~ EARTH'S ROT/lrtON
EARTH's
AXIS
Figure 4-6 The Coriolis effect in the eastward and
westward motion in the northern
hemisphere.
4-6.2 Geostrophic Inertial Components
4-6.2.1 Let us consider fluid particle A at a latitude <1>.
The horizontal velocity component U has a southerly
component Us and an easterly component U E. The vertical
component is W (Fig.4-7). Us may be resolved into Us sin <1>,
along a perpendicular AB to the earth rotation axis NS,
and a component Us cos <I> parallel to the earth's axis and
directed toward the south. The vertical component W,
can also be resolved into a component W cos <1>, along AB,
and a component W sin <I> parallel to NS and directed
toward the south. The components Us cos <1>, W sin <I>, being
parallel to the earth's axis, do not contribute to modifying
the distance of the fluid particle from the earth's axis, and
therefore, do not result in a Corio lis effect. The components
U , Us sin <1>, and W cos <I> lie in a plane which is perpen-
dicular to the earth's axis, and therefore contribute to
deflect the path of the fluid particle.
 Part 1 : Establishing the Basic Equations that Govern Flow Motion

N
alone, the fluid particle A has a velocity r(J) toward the
east. ((!) is the angular rotation of the earth.) The velocity
components relative to the earth are U E = r(d()jdt) toward
the east, and (U 8 sin <I>- W cos <I>)= drjdt along BA.
Therefore, the resultant velocity has an easterly component
r[(J) + (d()jdt)] and a radial component drjdt.
us
4-6.2.3 Consider a vector defined by the two components
X = R cos a and Y = R sin a. The derivatives of this vector
with respect to time are given by the derivatives of its
components, i.e.,
dX dR . da
- = - cos a - R Sill a -
dt dt dt
N dY dR . da
- = - Sill a + R cos a --
dt dt dt
The derivative of this vector is thus given by a sum. The
B first part, in the direction of the vector, has the magnitude
dRjdt. The second part has a magnitude R(dajdt).
Since

-sin a= cos (a+~) and cos a= sin (a+~).

the second part is oriented perpendicular to the vector
(X, Y) and directed toward the left of this vector.
We are concerned with the vector components in a
s plane X BY, perpendicular to the earth axis NS. It is seen

Figure 4-7 Velocity components of motion.

Figure 4-8 Plane perpendicular to the earths axis.
B
4-6.2.2 Let us now consider the components of motion
perpendicular to the earth's axis, U E Us sin <1>, and W cos <1>.
This is equivalent to considering a particle A which moves
in a plane perpendicular to the earth's axis. This plane
cuts the earth's axis NS at B, and let BA = r (Fig. 4-8).
Let M be a fixed point on this plane, which rotates with the
earth. The position of A is defined by the latitude <1>, BA = r,
and the angle MBA = e. It is assumed that () is positive
when A is to the east of M. Owing to the earth's rotation M
 Chapter 4: Inertia Forces

N B
z 47

r~
dt

d2 r

X
dt2
M Figure 4-10 Components of acceleration.
dr s
dt in the easterly direction, and a radial component equal to
Figure 4-9 Components of the derivative of a
vector r. r(w + ~~y
The total acceleration is composed of these four com-
from Fig. 4-9 that the component r(dejdt), perpendicular to ponents, as shown on Fig. 4-10.
the vector (X, Y) is actually to the east of this vector. If the radial component r[w + (dejdt)] 2 is written as
r[w 2 + 2w(d8jdt) + (dejdt) 2 ], further observations can be
4-6.2.4 The previous results will be applied to both the made.
radial component, dr/dt, and the easterly component, 1. The terms which are independent from w
r[w + (dejdt)], of the velocity vector. The acceleration
or time rate of increase of drjdt has a component d2 r/dt 2
in the same radial direction which is due to centrifugal
~;~ 2e:)(~~) + r ~:~ and -r(~~y
acceleration, and an easterly component perpendicular are the acceleration of the particle relative to the earth.
to BA which is equal to 2. The term rw 2 , directed toward B, is the centrifugal
acceleration.
dr
dt
(w +de)
dt
3. The two terms 2w(drjdt) and 2wr(dejdt) are per-
pendicular to and to the left of the velocity com-
ponents
Similarly, the acceleration or time rate of increase of the
easterly component {r[w + (dejdt)]} has a component, dr . --" de --"
dt = Us Sill 'I' - W COS 'I' UE and r-=
dt
respectively and are the geostrophic acceleration
terms.
Part I : Establishing the Basic Equations that Govern Flow Motion
4-6.2.5 The corresponding inertial components are
attained by multiplying these accelerations by the density
p. Therefore the corresponding geostrophic inertial forces
are
2pw(U s sin <I> - W cos <I>)
directed horizontally toward the east, and
2pwU E
directed along BA. Now, (2pwU E) has a horizontal com-
ponent (2pwU E sin <I>) and a vertical downward component
(2pw U E cos <I>). U E + Us = U is the vector sum of the
velocity components. The vector sum of (2wU s sin )
and (2wU E sin ) is (2wU sin ) along a horizontal per-
pendicular to the direction ofU and to the left ofU. Finally,
the geostrophic inertial force is composed of three terms:
1. A horizontal component 2pwU sin <I> perpendicular
to and to the left of U. In the southern hemisphere,
is negative, and the horizontal component
2pwU sin <I> is directed to the right of U.
2. A horizontal component 2pw W cos <I> directed toward
the west (due to the negative sign appearing in the
previous expression), and generally negligible as
the motions are generally so nearly horizontal that
W /U is very small.
3. A vertical component 2pwU E cos <I> directed down-
ward (or upward if the sign of U E is negative) which is
added or subtracted to the gravity, and which is
also negligible compared to gravity.
4-6.3 Total Inertial Components for a
Nearly Horizontal Motion
In the case of nearly horizontal motion, only the term
(2pwU sin <I>) is to be considered. n = 2w sin <I> is the
Coriolis parameter. Let us now have another two-dimen-
sional axis system, X, Y, fixed with respect to the earth's
surface and horizontal. The velocity components of U
are defined as u and v, respectively. Since the Coriolis
force acts perpendicularly and to the left of U, the two
geostrophic terms in this system of coordinates are:
- nv along the 0 X axis and + Qu along the 0 Y axis.
Adding the local and convective forces, with the vertical
component w neglected, the total inertial forces are
au + u-
p( -
au - nv )
au + v-
at ax ay
along the 0 X axis, and
av av av )
p ( -+u-+v-+Ou
at ax ay
along the 0 Y axis.
In the case of a periodic motion of frequency k, the local
acceleration term has an amplitude kU 0 , U 0 being the
maximum horizontal velocity. The Coriolis acceleration
amplitude is nu 0 . Therefore, it appears that the Corio lis
effect is important when Qjk is large, i.e., when the wave
period is of the order of magnitude of a day, which is th'e
case of tidal motion and storm surge.
PROBLEMS
4.1 Consider an unsteady two-dimensional flow where the
velocity components of x = L, y = L at time t = 0 are
u = L/t 0 , v = 2L/t 0 and at time t = t 0 are u = 2L/t 0 ,
v = 3L/t 0 .
Moreover, at time t = 0, the velocity components at
point x = 2L, y = L, are u = 1.2L/t 0 and v = 2.4L/t 0 and
at point x = L, y = 2L they are u = 1.1L/t 0 , v = 1.8L/t 0 .
Calculate the value of the total acceleration at x = L,
y = L by assuming that the variations of velocity with
time and distance are linear.
4.2 Calculate the total variation of temperature of a train
which travels 300 miles a day in the northern direction.
The mean daily variation of temperature is - 2F per
1000 miles. The daily variation at a given location is
4 sin (2nt/TtF where T = 24 hr.

4.3 In the case of a progressive acoustic wave in a pipe,
such that
. 2n
u = a sm L (x - Ct)
calculate the ratio of convective inertia to local inertia.
4.4 Consider a basin such as shown in Fig. 4-11, where the
particle velocity at the orifice is V = (2g1J) 112 , z = '1 is the
height of the free surface above the orifice. The horizontal
area of the basin being A = 4 ftl, (3716 cm 2 ), determine:
1. The variation of the free surface with respect to time,
i.e., the function l](t). At timet = 0, '1o = 20ft (609 em).
The local inertia will be neglected for this calculation.
2. The local inertia at point x = 3 ft (91.44 em) z = 0,
i.e., in the converging section.
3. The convective inertia at the same point.
4. Is the neglect of the local inertia a valid assumption?
Explain. The friction will be neglected.
Figure 4-11
Chapter 4: Inertia Forces
5. When a constant discharge q0 = 1 fe /sec (28,317 49
em 3 /sec) is poured into the tank, establish the function
IJ(t) ('1o = 20ft, 609 em) and determine '1 when t ~ oo.
The discharge coefficient Cn will be assumed to be unity.
4.5 Determine the convective inertia terms which can be
neglected in a jet.
4.6 The influx of discharge into a reservoir is defined by
the equation
Q(t) = 10,000 (1.5- sin kt) ft3/sec
= [283 (1.5- sin kt) m 3 jsec]
where k = 2n/T and Tis a period of 1 year. The horizontal
area of the reservoir is defined by
A(z) = 10,000z2 ftl (929z 2 m 2 )
The top of the spillway for flood discharge is located at an
elevation z 1 = 100 ft (30.48 m) and has a discharge capacity
Qs = C1[z - z 1] [2g(z - z 1 )] 112
where the coefficient of discharge C = 0.5, and the length
of the spillway 1 = 100 ft (30.48 m). The turbined discharge
is constant and equal to 7000 ft 3 /sec (198m 3 /sec).
Determine the variation of the level of the free surface in
the reservoir as a function of time and the maximum
discharge over the spillway for each year following time
t = 0. The maximum possible discharge over the spillways
will also be determined. At time t = 0, one will take the
free surface elevation at z = 30ft (9.144 m).
4.7 Consider a periodic two-dimensional oscillation in a
rectangular tank (seiche) of length 1 and depth d(l ~ d)
(Fig. 4-12). The period of oscillation is T = 21/(gd) 112
The horizontal velocity component u is assumed to be a
constant along a vertical and is a function of time only.
The equation of the free surface is (a ~ d):
h = d + a cos mx cos kt
m = 2n/L 21 = L = T(gd) 112 k = 2n/T
 Part 1 : Establishing the Basic Equations that Govern Flow Motion

50 where W is the vertical velocity component, U E is the east-

west velocity component positive to the east, Us is the
north-south velocity component positive to the north,
r is the distance from the center of the earth, <I> is the latitude,
and () is the longitude.

4.12 Compare the vertical component of geostrophic

acceleration with the gravity, in the case where U E cos <I> = 1
Figure 4-12 knot (1.688 ftjsec: 51.4 em/sec).

Determine the maximum value of u and w and the location

where they are maximum. Some simplifying assumptions
will be accepted for these calculations. Determine the
expression for the local inertia and the convective inertia,
their maximum values, and the ratio of the maximum
value of convective inertia to the maximum value of the
local inertia. Present a criterion permitting the evaluation
when the convective inertia is negligible.

4.8 Demonstrate that

-du = -ou + 2(w'7 - vO + -o -

ox
(V2 2
)
dt ot
and find similar expressions for dvjdt and dw/dt.

4.9 Express the components of acceleration A,, A 6 , A= in a

cylindrical system of coordinates.

4.10 Define the components of velocity u, v, w in a natural

system of coordinates, i.e., the axes 0 X and 0 Y are defined
from a given point along and perpendicularly, respectively,
to a streamline. Give the components of acceleration. In a
second case, where it is assumed that paths and streamlines
are different, the axis will then be defined at a given point
with respect to paths. Then, give also the components of
acceleration (R is the radius of curvature of the path).

4.11 Demonstrate that in a spherical system of coordinates

(r, e, <I>) related to the earth, the operator djdt is:

d o o UE o Us o
-=-+
dt dt
W-+---+--
or r sin <I> oO r o<l>
 sr
Chapter 5 5-l Internal and External Forces
The applied forces on an elementary mass of fluid
consist of internal forces and external forces.

Applied Forces 5-1.1 Internal Forces

Internal forces result from the interaction of the interior
points of the considered mass of fluid. According to the
principle that action equals reaction, these internal forces
balance in pairs and their sum is zero. Their total torque
is also zero. However, the work of these internal forces is not
zero. It is for this reason that it is important to mention
their existence. For example, the headloss in a pipe is the
result of the work of internal viscous forces.

5-1.2 External Forces

The forces on the boundaries ofthe considered particle of
fluid, called surface forces, and the forces which are always
acting in the same direction on its mass, called body or
volume forces, are not balanced. These are the external
forces.

5-1.2.1 Surface forces result from forces acting on the

outside of the considered volume. They are caused by
molecular attraction. They decrease very quickly away
from the boundaries of the particle of fluid, and their action
is limited to a very thin layer. In practice, if the fluid is a
continuous medium, this layer can be considered infinitesi-
mally thin and blended with the surface of the particle.
These surface forces can be divided into (1) normal forces-
due to pressure; and (2) shearing forces-due to viscosity.
These two kinds of forces also exist within the particle, but
are always balanced in pairs and their sum is zero, as
previously noted.

5-1.2.2 Body forces result from an external field (such

as gravity or a magnetic field) which acts on each element
of the considered volume in a given direction. For this
reason, they are called body or volume forces. Except for
some rare cases, for example the study of the motion of a
 Part 1 : Establishing the Basic Equations that Govern Flow Motion

52 fluid metal in a magnetic pump, only gravitational force

has to be considered in fluid mechanics. In general, this
gravitational force can be considered as acting in the same
fixed direction. However in some studies, such as tidal
motion and oceanic circulation, the gravity acceleration
must be considered radial.
5-1.2.3 Other applied external forces are:
1. The capillary forces due to the difference of molecular
attraction between two media are of particular
importance in free surface flow through a porous
medium and in small gravity waves such as generated
by a gentle breeze. These are called capillary waves.
2. The geostrophic force caused by the Coriolis accelera-
tion due to the earth's rotation is sometimes con-
sidered as a body force similar to the gravity force,
even though it is really an inertial force.
5-2 Gravity Forces
Similar to the inertia forces, the volume forces are pro-
portional to the mass of the fluid and to the acceleration
caused by an external field. In the case of gravity action,
the volume force per unit of volume is simply equal to the
fluid weight: 05 = pg, where g is the acceleration due to
gravity. This force is independent of the motion. It is the
same whether the fluid is static or in any viscous or turbulent
motion.
The components X, Y, Z of the gravity force, expressed
in a differential form in the three-axis system OX, OY, OZ
are given below. The vertical axis OZ is considered as
positive upward along the normal to the earth's surface.
The components are X = 0; Y = 0; and Z = - pg =
5-3 Pressure Forces
Pressure forces result from the normal components of
the molecular forces near the boundary of the considered
volume. The magnitude of the pressure at a point is obtained
by dividing the normal force against an infinitely small area
by the area.
5-3.1 Pressure Magnitude, Pressure
Force, and Direction
The pressure magnitude is a scalar quantity that is
independent of the orientation of the area on which the
force is applied. This may be demonstrated by considering
a triangular two-dimensional element in a fluid at rest
(Fig. 5-l).
Since there is no motion, inertia forces and viscous forces
are zero, and the only forces are gravity and pressure. The
projections of these forces along the 0 X and 0 Y axes
yield the equalities
Px dy - p ds sin rx = 0
dxdy
Py dx - p ds cos rx = pg - 2-
Introducing dy = ds sin rx, dx = ds cos rx, and neglecting
the second-order term, pg(dx dy/2), one obtains p = Px,
p = Py Hence, p = Px = Py Since rx is any arbitrary angle,
Figure 5-1 Pressure magnitude is independent
of orientation.
y

L.
-ojoz(pgz). In vector form, they are -grad(pgz), since
X= -o/ox(pgz) = o and Y = -o/oy(pgz) = o.
The gravity force is often neglected in gas dynamics
except, for example, in meteorology or in the calculation
for chimneys and ventilation openings, where the pheno-
mena are influenced by the variations of gravity forces
due to density changes.

the pressure is seen to be the same in all directions. A
similar demonstration is possible in three dimensions.
However, it is evident that the gradient of pressure force
(which is a vector) changes with direction. In the same way,
the force caused by pressure against an area (which is a
vector) changes direction as the normal to the considered
area changes direction.
5-3.2 Rate of Pressure Force Per Unit
of Volume
Consider an elementary fluid particle (dx dy dz) (Fig. 5-2).
The pressure force due to the external adjacent fluid
particle acting against the side ABCD is p(area ABCD) =
p dy dz. The pressure force against the other side acts in
the opposite direction and may be written:
- (P + ~~ dx) (area EFGH) = - (P + ~~ dx) dy dz
Hence, the difference of pressure forces acting in opposite
directions is
ap ) ap
p dy dz - ( p + ax dx dy dz = - ax dx dy dz
Similarly, the pressure force differences acting in the
OY and OZ directions are -(apjay) dx dy dz and
-(ap;az) dx dy dz.
Figure 5-2 Difference in pressure forces in a unit
of volume.
p
Chapter 5: Applied Forces
Hence, the rate of change of pressure force per unit of 53
volume is given by the three components -ap;ax, -apjay,
and - apjaz, which can be written vectorially: -grad (p).
5-3.3 Fluid Motion and Gradient of
Pressure
It is interesting to note that the motion of the fluid
particle does not depend upon the absolute value of p, but
only upon the gradient of p. Consider the motion in a
tunnel. The motion depends upon the difference in pressure
levels between the upstream and the downstream sections.
Therefore, it is possible to operate a scale model experiment
at any convenient arbitrary absolute pressure, provided
the pressure gradient is kept in similitude. However, if the
pressure level drops below a critical value, cavitation occurs
and, for similitude, the scale model must be operated
under partial vacuum (see Appendix B).
5-3.4 Pressure and Gravity
The total force due to the pressure force and gravity force
per unit volume is
grad p + grad pgz = grad (p + pgz)
The sum of these two linear quantities (p + pgz) is a
constant in hydrostatics since p - Pa = - pgz where Pais a
constant external pressure (atmospheric). This property is
also verified in a cross section of a uniform flow as in a
channel or in a pipe, or more generally when the curvature
of the paths is negligible or the motion is very slow (see
Section 10-2.1.1). Hence the sum (p + pgz) may often be
conveniently replaced by the single term p*: p* = p + pgz.
In hydrostatics p* =constant. Whereas pjpg is known as
the pressure head, p* / pg is called the piezometric head.
5-4 Viscous Forces
5-4.1 Mathematical Expressionfor the
Viscous Forces
Shear stresses are present because of fluid viscosity and
are caused by the transfer of molecular momentum.
The friction force r is assumed to be proportional to the
 Part 1 : Establishing the Basic Equations that Govern Flow Motion

G H
54 More generally, for a three-dimensional incompressible
fluid, it is possible to demonstrate that the friction force
u+~dy components per unit of volume are:
oy
y
c 02U 02U 02U)

L.
D
J1 ( ox2 + oy2 + oz2 = J1'V2u

o2v o2v o 2v)

J1 ( OX2 + oy2 + oz2 = J1'V2v
A u 8
o 2w o 2w o2w)
J1 ( OX2 + oy2 + oz2 = J1'V2w

They are written vectorially:

o2 V o2 V o2 V)
E F

Figure 5-3 Two-dimensional element offluid. J1 ( oxz + oyz + ozz = Jl'VzV

5-4.2 Approximations Made on Viscous

coefficient of viscosity J1 and to the rate of angular deforma- Forces
tion.
It has been shown experimentally that it is sometimes
Consider a two-dimensional element of fluid (Fig. 5-3).
possible to consider friction effects as negligible. The
The friction force on the side AB of length dx is: r dx =
preceding expression for the friction forces show that they
Jl(oujoy) dx. Since the velocity at C is (u + (oujoy) dy), the
can be neglected when the Laplacian of the velocity
friction force on the side CD is
components (V 2 V) is small. This is often true outside the
boundary layer where the fluid motion is similar to that of a
(r + ~: dy) dx = J1 :y ( u + ~; dy) dx perfect fluid.
Sometimes it is possible to neglect only one part of the
au ou
2
viscous friction terms. For example, as similarly explained
= J1 oy dx + J1 oy2 dy dx.
in Section 4-5.2.4, in a two-dimensional laminar boundary
These forces act in opposite directions. If the force due layer or in a jet, o2vjox 2 and o2vjoy 2 and also o2ujox 2
to the particle GHCD acts in the OX direction on the side may be neglected in comparison to o2u/ol, and only the
CD of ABCD, the force due to the particle ABCD will act friction force J1(o 2u/oy 2) has to be taken into account.
in the same direction of the AB side of a ABFE and by
reaction ABFE will cause a force in the - 0 X direction on 5-5 Some Theoretical Considerations
ABCD. The total shear force thus becomes:
of Surface Forces
or o2u
oy dx dy = J1 ol dx dy 5-5.1 A General Expression/or Surface
Forces
Dividing by dx dy, the friction force per unit of area is:
Surface forces, as previously seen, consist of pressure
or o 2u force and friction force. These surface forces may be intro-
-=jl-
oy oy 2 duced without consideration of their physical nature. The
 Chapter 5: Applied Forces

advantage of so expressing the surface forces lies in its Hence, the external forces may be defined by a tensor of 55
applicability to any kind of motion, e.g., perfect, viscous or rank two:
turbulent, compressible or incompressible. However, the
values of these surface forces are expressed differently when
their physical nature is taken into account.

5-5.2 The Nine Components of the

External Forces: Components of Lame These forces are completely given by Table 5-l. Now,
the addition of all the forces per unit volume acting in the
5-5.2.1 Consider an elementary mass of fluid in the form same direction yields
of a cube; its edges are parallel to the three coordinate
axes OX, OY, and OZ, as shown in Fig. 5-4. In the 0 X direction:
On each side of this elementary cube, surface forces may
be completely defined by three components parallel to the
three coordinate axes. Two of these components are shear
stresses while the third is a normal stress. In the 0 Y direction:
Since a cube has six sides, 18 components have to be con-
arxy a(Jyy arzy)
sidered. These components are defined with the help of two (- + - + -
subscripts. (J are the normal forces and r are the shearing ax ay az
forces. The first subscript x, y, or z refers to the axis normal In the 0 Z direction:
to the considered side. The second subscript x, y, or z
refers to the direction in which the force acts.
The pairs of parallel forces acting on two opposite sides
of the cube act in opposite directions, and their difference is
obtained by a simple partial derivative with respect to the 5-5.2.2 On the other hand, consider the torque of a fluid
distance between the two considered sides. particle about one point (for example, A in Fig. 5-5).

SHEARING
FORCE
/
/
T XZ /

I OT xy
NORMAL r,Y + ~ dx
FORCE
I
I
Figure 5-4
I I
Notation for surface forces. I
lo
}-':--- -
/
/
/

A E

y OT xz
Txz + ~dx
 Part 1 : Establishing the Basic Equations that Govern Flow Motion

s6 Table 5-1 External forces

Stresses applied to the side normal to the axis

ox OY oz
On the side of area { ABCD EFGH DHGC AEFB AEHD BFGC
(see Fig. 5-4) dydz dy dz dxdz dx dz dx dy dx dy

ox ~XX
( ~XX+ a~xx dx )
a;: Tyx (
Tyx
oryx )
+ aJdy '=x
(
r zx + ---a;
OTzx
dz )

In the OY Txy ( Txy + ---a;

orxy dx ) ~yy (~ yy + a~yy
ay
dy) Tzy ( Tzy orzy )
+ Tzdz
direction of

oz Lxz (
r xz + ---a;
OTxz
dx ) Tyz ( Tyz oryz )
+ aJdy ~zz
(
~ ==
a~zz dz )
+ ---a;

The total sum of the torque caused by the shearing
stresses is
rxy(dy dz) dx - ryx(dz dx) dy
This torque is equal to the mass times the square of the
radius of gyration (dRf, times the square of the angular
velocity (w 2 ), which may be written p dx dy dz(dR) 2 w 2
Since dR is infinitesimally small, having the same order
as dx, dy, and dz, (dR) 2 is of the second order and the speed
of gyration becomes infinite, which is physically impossible.
Hence, the total torque must be zero. This condition is
possible only when rxy = ryx
Figure 5-5 Torque about the point A.
Similarly, it may be shown that ryz = rzy and rxz = rzx
Hence, the nine components of the tensor of external force
are reduced to the six so-called components of Lame.
5-5.3 Value of the Lame Components in
Some Paracular Cases
5-5.3.1 In the case of a perfect fluid, the shearing stresses
are zero and the normal forces become simply the pressure
forces:
(Jxx = (JYY = (Jzz = - P
rxy = !yz = !zx = 0

5-5.3.2 In a viscous incompressible fluid, it is possible to

demonstrate that the normal forces (CJ) are the sum of the
pressure force and a viscous force proportional to the
coefficients of linear deformation:

+ 2f.1-
au
== -p
(J XX
ax
Similar expressions exist for CJ YY and CJ ==.
 Chapter 5: Applied Forces

The shearing stresses r are functions of the coefficients The work of all the forces in the 0 X direction is: 57
of angular deformation:
oa xx ) dydz ( u+ oxdx
(Jxxdydzudt- ( (Jxx+~dx au ) dt

!yz = (aw + OZav)

11 oy
+ !xy dy dz U dt - ( !xy + o;;y dx) dy dz ( U + ~~ dx) dt
and similar expressions for rxz and rxy
Now, introducing these values in the sum of forces
acting in the same direction as, for example, in the 0 X + !xz dy dz U dt- (rxz + o;;z dx) dy dz (u + ~~ dx) dt
direction, it is easy to verify that
The total work per unit volume changed into heat in a
oaxx +orxy unit period of time is called the "dissipation function" <1>.
( -- - +orxz)
- = --
op + 11vn l U It is a function of the linear and angular rates of deforma-
ax ay az ax
tion and is found by substituting the values of a and r.

+ 11[2(au)
ox + 2(av)
oy + 2(aw)
5-5.3.3 In the case of a viscous compressible fluid, the 2 2 2
shearing stresses are the same as in the above case, but the <l> = A(div V) 2 oz
normal forces have to take into account the change of
volume of the fluid particle. It may be seen that: + (aw + av) 2 + (au + aw) 2 + (av + au) 2 ]
oy az az ax ax ay
au av ow) au
axx = - P + A ( ax + ay + az + 211 ax A(div V) 2 is equal to zero in an incompressible fluid.
This function can be used, for example, in the calculation
of head loss of a viscous flow in a pipe and the damping of
Two similar relationships are easily deduced for ayy and
gravity waves.
a zz- A is a second coefficient of viscosity for a gas. From the
kinetic theory of gases, it may be shown that for a mono-
atomic gas 3A + 211 = 0. In practice this relationship is
considered accurate enough for any kind of gas. PROBLEMS

5.1 Demonstrate that the viscous forces acting on an

5-5.4 Dissipation Function element of incompressible fluid of volume unity can be
expressed in terms of the rotation by the following expres-
The energy transformed into heat either by change of sion:
volume or by friction may be obtained by adding the work
done by all the external forces. This work is equal to the _ 211 (a' _ary)
oy az
external forces times their displacement.
For instance, in the OX direction, the work of pressure The two other expressions are obtained by permutation.
forces is:
5.2 Demonstrate that in an irrotational flow of an incom-
p dy dz u dt - (v + ~~ dx) dy dz ( u + ~~ dx) dt pressible fluid, the sum of the viscous forces is theoretically
zero.
 Part 1 : Establishing the Basic Equations that Govern Flow Motion

58 5.3 Calculate the viscous force acting on a cubic element of

water of volume 10- 3 L 3 and located between y = / 0 L and
y = / 0 L in a two-dimensional flow defined by the velocity
components u = 10- 4 (gjv)(4 - y)y, v = 0, and w = 0. Cal-
culate the values of a and r acting on each side of this cube,
and the rate of dissipation of energy in the cube. (v =
1.076 X 10- 5 ftljsec; 0.01 cm 2 /sec.)

5.4 Express J.N 2 V in a cylindrical system of coordinates.

5.5 Express the stresses a and r in a cylindrical system of

coordinates for an incompressible fluid.
 Chapter 6 6-1 Main Differential Forms of the
Momentum Equation
59

The momentum equation is obtained by equating the

applied forces to the inertia force for a unit volume of the
Forms of the fluid. The physical meaning and the mathematical expres-
sions of these forces have been developed in Chapters 4
Momentum Equation: and 5.
The different forms of the momentum equation cor-
Equations ofEuler responding to a number of cases encountered in hydro-
dynamics are now presented.
and Navier-Stokes 6-1.1 Perfect Fluid: Equations of Euler
6-1.1.1 The first major approximation is to assume that
the fluid is perfect. In this case the friction forces are zero
and the applied forces consist of gravity and pressure only.
The momentum equation is obtained directly from the
expressions developed in Chapters 4 and 5, in the three-axis
system OX, OY, OZ, where OZ is assumed to be vertical
(see Table 6-1 ). When the expressions dujdt and p* are

Table 6-1 The momentum equation

Inertia forces Pressure and gravity forcest

per unit of volume per unit of volume of fluid
(see Section 4-1.1) (see Section 5-3.4)

ap*
ax
8p*
ay
dw 8p*
pdt az
Written in vector notation, these become
dV
p dt + grad p* = 0

t Recall (p* = p + pgz).

 Part 1 : Establishing the Basic Equations that Govern Flow Motion

6o expanded (see Section 4-4.1), the momentum equation Two similar equations give the value of opfoy 0 and opfoz 0 by
takes the form along the OX axis given by Equation 6-1. permutation of x 0 , y 0 , z 0 , which are the coordinates of a
particle at time t = t 0 . These are called the equations of
Inertia forces Applied forces Lagrange.

Local Convective Pressure Gravity

inertia inertia 6-1.2 Viscous Fluid and the
Navier-Stokes Equations

(-
p
ou) =
ou + w-
ou + v-
au + u- a
--(p + pgz) (6-1)
at ax oy az ax
Two similar equations may be written in the OY and OZ
directions. These are called the equations of Euler.
Such a system of equations associated with the continuity
relationship oufox + ovfoy + owfoz = 0 forms the basis
of the largest part of the hydrodynamics dealing with a
perfect incompressible fluid. These equations are mathe-
matically of the first order but are nonlinear (more specif-
ically quadratic) because of the convective inertia terms.
This quadratic term is the cause of most mathematical
difficulties encountered in hydrodynamics.
6-1.1.2 It has been explained in Chapter 1 that it is
possible to study hydrodynamic problems either in Eulerian
coordinates or in Lagrangian coordinates. It is recalled
that the Lagrangian method consists of following particles
along their paths instead of dealing with particles at a
given point. This method is used, for example, in some
studies of periodic gravity waves over a horizontal bottom.
The corresponding equations are given here only for the
purpose of recognition in the literature and will not be
developed.
If X, Y, Z are the volume or body forces, i.e., gravity,
the Lagrangian equation along the OX axis is written:
6-1.2.1 If the friction forces are introduced in the Eulerian
equations, the Navier-Stokes equations are obtained (see
Section 5-4.1), as shown in Equation 6-2.
The Navier-Stokes equations are the basis of most
problems in fluid mechanics dealing with liquid. They are
second-order differential equations because of the friction
terms, and nonlinear because of the convective inertia
terms.
6-1.2.2 These Navier-Stokes equations are written in a
very concise manner with the aid of tensorial notation.
Although a knowledge of tensorial calculus is not required,
it is given here as a guide to further reading on the subject.
Use is made of two subscripts, i and j, which indicate
when an operation is to be systematically repeated and
which component of a vector quantity (such as V) is being
considered. When an index is repeated in a term, the
considered quantity has to be summed over the possible
components. For example, the continuity equation
oufox + ovfoy + owfoz = 0 is tensorially written: ou;/ox;
= 0, since the subscript i indicates that the quantity (here
V) has to be summed over the three components OX,
OY, OZ.
The three previous Navier-Stokes equations, may be
written simply as:

Here, the subscript i is called "free index" and indicates the

component being considered; the subscript j, called
"dummy index," indicates repeated operations.
 Chapter 6: Forms of the Momentum Equation: Equations of Euler and Navier-Stokes

61
Inertia forces Applied forces

Local Convective Pressure Gravity Friction

inertia inertia

op
ox

ov ov ov ov) op
(
p~+u~+v~+w~ = (6-2)
ot ox oy oz oy

p( 8t
OW OW
+ u OX + v oy + w
OW OW)
oz = -
o(p +
oz
pgz)
+ f.1
(oox
2w
2 +
o2 w
oy 2 +
o 2 w)
oz 2

6-1.2.3 These Navier-Stokes equations are often written Since the derivative of the sum in parentheses is zero in all
in another way in order to emphasize the role of the directions, one obtains
rotational component of motion. It is sufficient in this case
to use the expression of the inertia force demonstrated
v2
pT + p + pgz = constant
in Chapter 4, which yields (see Section 4-4.3) Equation 6-3.
which is the well-known Bernoulli equation, fully developed
6-1.2.4 The three components of Equation 6-3 are more in Chapter 10.
concisely written in the vector form of Equation 6-4, which
may be transformed as 6-1.3 The General Form of the
Momentum Equation
grad ( pv- + p + pgz)
2
= -
ov
p - - p(curl V) x V + f.1V 2 V It has been shown that the applied forces may be ex-
2 ot pressed independently of their physical nature with the help
of the tensor of rank two:
6-1.2.5 In the case of a steady (oV jot = 0) irrotational
flow (curl V = 0) of a perfect fluid (f.l = 0), the above
equation gives at once:

v2
grad ( pT+ p + pgz )=o The main advantage of such a notation is that it is valid
for any kind of fluid-perfect or real-and any kind of
Part 1 : Establishing the Basic Equations that Govern Flow Motion

Inertia forces Applied forces

Local Convective Pressure Gravity Friction

inertia inertia

Caused by Caused by
variation rotation
of kinetic
energy

p[ -ou
ot
(V
+ -o -
ox 2
2
) + 2(w1J J
- v() = - o(p + pgz) +
OX
n2
p,v U

p[ -ov
ot
+ -o -
oy
(V2 2
) + 2(u( - w~) J= - o(p +
oy
pgz) nz
+ p,v v (6-3)

p [ -ow
ot
(V
+ -o -
oz 2
2
) + 2(v~ J
- U1J) = - o(p
+ pgz)
OZ
n2
+ p,v W

Inertia forces Applied forces

Local Convective Pressure Gravity Friction

inertia inertia

Caused by Caused by
variation rotation
of kinetic
energy

p[~~ +grad (~ 2) +(curl V) x J

V =-grad (p + pgz) + p,V 2 V (6-4)
 Chapter 6: Forms of the Momentum Equation: Equations of Euler and Navier-Stokes
Inertia forces Applied forces
~
Volume
forces
dv
p dt
motion-laminar or turbulent. It will be shown that if
in the momentum equation the real values u, v, w, and p
are replaced by the average values ii, v, w, and p in a tur-
bulent flow, the surface forces (J and r include additional
components due to the turbulent fluctuations.
Hence, the advantage of using the notations (J and r
exists in expressing general equations which are independent
of the nature of the flow. Equating the inertia forces to the
applied forces expressed in the manner shown in Chapter 5
yields Equation 6-5.
In practice, if OZ is vertical upward, X = 0, Y = 0,
z = - pg = -(ojoz)(pgz).
6.1-4 Synthesis of the Most Usual
Approximations
Tables 6-2 and 6-3 recall the physical meaning of
different terms and possible approximations accepted
in the studies of flow motions, which may be investigated
in the following. Complex disordered and random motions,
even though also obeying the Navier-Stokes equation,
cannot be analyzed on a purely Newtonian deterministic
Surface
forces
(6-5)
approach. The motion is averaged and the friction term
J1'V 2 V is replaced by an empirical functional relationship
proportional to V in the case of flow trough porous
medium, or to V 2 in the case of fully turbulent motion.
These two points are analyzed in Chapters 8 and 9.
6-2 Exact Integration vs Numerical
Solutions
6-2.1 An Example of an Exact Solution
of Navier-Stokes Equations. Flow on a
Sloped Plane
6-2.1.1 It is to be expected that a general solution of the
system of differential equations given by the continuity
and momentum principle does not exist. However, some
exact solutions can be obtained if the boundary conditions
are simple. Examples where exact solutions may be obtained
include flow between parallel plates (i.e., the Couette
flow, the Poiseuille flow), flow due to a rotating disk,
 Part I : Establishing the Basic Equations that Govern Flow Motion

64 Table 6-2

[av
Par + grad(~2 ) + (curl V) x VJ - grad(p + pgz) + J.i'V 2 V

Physical Local Variation Rotational Pressure Gravity Friction

meaning inertia of kinetic term force force force
energy
with space

Convective inertia Applied forces

Mathematical First-order Nonlinear (quadratic) term First -order Constant term Second-order
characteristics linear linear linear
term term term

Approximation =0 In a =0 for ==o in a gas =0 in an

steady flow irrotational (with exceptions) ideal fluid
motion;
solution given
by a harmonic
function

=0 =0
for slow motion

uniform unsteady flow over an infinite flat plate (see The Navier~Stokes equations are reduced to:
Section 13-2.3.3), etc.
pg sin a+ J!G:~) = 0
6-2.1.2 The very simple example of a two-dimensional
steady uniform flow on an inclined plane of infinite dimen-
ap
0 = - - - pg cos a
sions is given here as an example (Fig. 6-1); the Navier~ ay
Stokes equation given in Section 6-1.2.1 may be simplified The second equation yields
in the following manner: p = Pa - pgy cos a,
Since the motion is steady, aujat = 0 and avjat = 0.
Since the motion is two-dimensional, w = 0, and all where p. is the atmospheric pressure. Hence the lines of
derivatives with respect to z are zero. Since the motion is equal pressure are parallel to the OX axis.
uniform and parallel to the axis OX, v and all its derivatives The boundary conditions are u = 0 for y = - d on the
are zero. All derivatives with respect to x are also zero. plane, and dujdy = 0 for y = 0 at the free surface. Taking
The components of the gravitational force are X = pg sin a into account these boundary conditions, the integration of
and Y = - pg cos a. Since the flow is uniform, v = 0, and a2 u = - -sma
-
pg .
the continuity equation is reduced to aujax = 0. ayz Jl
 Chapter 6: Forms of the Momentum Equation: Equations of Euler and Navier-Stokes

Table 6-3 6s

Convective
Local inertia inertia Friction Equations Some applications

Steady motion Slow Without friction grad (p + pgz) = 0 Hydrostatics

or motion con- motion With friction -grad (p + pgz) + JlV 2 V = 0 Steady uniform flow
sidered as a Flow in a porous medium
succession of
steady motions lrrotational Without friction Nonuniform (convergent)
2
motion grad (p ~ + p + pgz) = 0 Steady flow at a constant
total energy. Calculation
of pressure in a two-
dimensional flow net

Rotational With friction General case of steady

motion grad (p ~2 + p + pgz)) motion; laminar boundary
layer theories
= - p(curl V) x v + JlV 2 V
Unsteady motion Slow Without friction av Gravity wave (first-order
motion pat + grad (p + pgz) = 0 theory); water hammer
theory
With friction av o
Gravity wave damping
pat + grad (p + pgz) - Jl'\1 2 V =

Most nonlinear wave

av + grad ( p 2
Irrotational Without friction v + p + pgz) = 0
2
motion P at theories
Gravity wave theory of
Rotational
motion
Without friction
grad(p ~ 2 + p +pgz) Gerstner

av =0
+pat+ p(curlV) x V

With friction General case Tidal wave in an estuary

gives successively, The discharge per unit of width is:

OU g sin IX
-=---y
ay v
q= Jo u dy
-d
= g sin
2v
IX Jo (d2 -l) dy
-d

and = g sin IX d 3
_ g sin IX (d 2 -y2)
U---
3v
2v
The loss of energy per unit length may be given by the
which is the equation of a parabola. dissipation function <1>, which in this case is J.l(oujoy) 2
 Part 1 : Establishing the Basic Equations that Govern Flow Motion
66
Figure 6-1 Laminar flow on an inclined plane.
Hence the loss of energy per unit length is
ld<l>dy = J1 Jd (au) dy = ..:._::__sin_oc) _d
-
2 (pg
o o ay 3J1
This can also be obtained by determining the work done
by friction forces Fr as follow;
s:Fr du = J1 s: :~ du = J1 s: G~Ydy
6-2.2 Numerical Treatments of the
Navier-Stokes Equations
It is now possible, thanks to the development of high-speed
computers, to treat the Navier-Stokes equations directly
by finite differences. This permits the study of complex
flow motions beyond the usual limits of analytical solutions.
Among many possible methods which have been
developed, one must mention the MAC (markers and cells)
method for two-dimensional or axially symmetric in-
compressible fluid, and the PIC (particle in cell) method
for two-dimensional compressible fluid.
In brief, these methods consist in solving time dependent
flow motion at successive intervals of time from a given
set of boundary conditions and the knowledge of the flow
motion at time t = 0. The space intervals define a square
mesh or a grid. Considering one (or two) particle(s) at the
center of each of these squares at time t = 0, it is then
possible to calculate the paths of these particles at successive
intervals of time. The results are printed directly by the
computer and give a Lagrangian representation of the
flow pattern as a function of time. It is also possible to
obtain and print velocity vectors and pressure distributions
(isobars) directly.
It is easily realized that this method is extremely powerful,
as is evident from Fig. 6-2. This figure represents the flow
patterns which will be obtained by the sudden release of a
vertical wall of water (this is the dam-break problem)
and hitting an obstacle. However, this method, as any
numerical method, also has its limitations. The accuracy
of the results is rapidly limited by the error which is made
by replacing differential terms by finite difference. These
are the truncation errors. A round-off error is also added,
2 3 as will be explained in further detail in Section 18-3.1.2.
Any calculation also requires a preliminary analysis
of stability conditions so that the cumulative error does
not increase out of proportion. This method- is costly
due to computing time. Nevertheless, it is to be expected
that these kinds of methods will be used more and more for
solving problems of increasing complexity. There are
theoretical limits, however, to this method, since the
phenomena of turbulence (which is three-dimensional)
could not be analyzed in a two-dimensional finite difference
scheme. Also, the viscosity coefficient J1 is taken arbitrarily
larger than its real value for the sake of numerical stability.
6-3 The Stability of Laminar Flow
6-3.1 The Natural Tendency for Fluid
Flow to Be Unstable
Consider two layers of fluid moving with different
velocities because of the effect of friction (Fig. 6-3). If for
any reason a small undulation exists between these two
layers, the velocity of layer 2 decreases; hence, according
to the Bernoulli equation, the pressure tends to increase.
On the other hand, the velocity of layer 1 tends to
increase; hence, the pressure tends to decrease. The pressure
action being in the same direction as the inertial forces
Figure 6-2
An example of an application of numerical treatment of the
Navier-Stokes equation. (Courtesy of Dr. F. Harlow.)
 Part 1 : Establishing the Basic Equations that Govern Flow Motion
68 INERTIA (CENTRIFUGAL) ACTION
PRESSURE ACTION
(I)
(2)
Figure 6-3 Fluid flow is fundamentally unstable.
(by centrifugal action), the undulation has a natural
tendency to increase in amplitude. However, this increase
in path length of the particles in motion causes an increase
in friction effect, which in turn has a tendency to dampen
such an undulation. This disturbance would tend to be
dampened out by friction unless there is a transfer of
energy (or a transfer of momentum by convective inertia
forces) from the primary motion to this disturbance. Hence,
in a turbulent motion, the rate of turbulence depends on
the rate of energy which is transmitted from the primary
flow to be finally absorbed entirely by friction.
6-3.2 Free Turbulence, Effects of Wall
Roughness
Instabilities initially occur either within the fluid or
at a boundary. In the first case the phenomenon is called
"free turbulence." The lumps of turbulence come initially
from the zone where the gradient of kinetic energy is a
maximum, as for example from the boundary of a submerged
jet. Hydraulic jumps, breaking waves on a beach, and
whitecaps generated at sea under wind action, are also
cases of free turbulence. Initial instabilities are most often
caused by roughness of a fixed boundary. Indeed, any
roughness causes a local increase of velocity which con-
sequently produces a local scrong gradient of kinetic
energy, causing an instability (Fig. 6-4). This instability
also arises between two fluids of different density. For
example, the wind blowing on a liquid causes ripples.
These ripples are due to an instability between the air flow
and water.
The very interesting question concerning the origin of
turbulence will not be studied in detail here. It is simply
emphasized that the Navier-Stokes equations give unstable
solutions which represent exact motions only at low
Reynolds numbers, i.e., when the friction forces are large
in comparison with the kinetic energy gradient. Turbulence
will be studied further in Chapters 7 and 8.
Since exact motions occur only at low Reynolds numbers,
the stability of laminar flow depends upon the ratio of the
gradient of the kinetic energy (dimensionally equal to the
convective inertia forces) to the viscous forces. This ratio
is dimensionally equivalent to the Reynolds number
(Re = VL/v), where Vis a velocity, and Lis a characteristic
length.
Because of the definite instability of the flow, a distur-
bance caused by external forces (such as that created by a
roughness at the boundary) grows exponentially if the
disturbance is large enough. If the disturbance is small,
the friction forces cause its damping. But if the ratio of the
gradient of kinetic energy (dimensionally equal to a
convective inertia force) to the viscous force is large enough,
even an infinitely small disturbance is able to cause in-
stability. Hence, laminar flow is naturally and basically
unstable at large Reynolds numbers. But even at low
------
Figure 6-4 Instability of a laminar flow because of
a roughness.
~
1 111171/lj)lllllll
-------------==
------~
___.....--- llA u~
..,.1..,1-,7-,l-1,..-,1;T''// II! II I l l /
~
I I I I I I I f/>.1 I I I I I II
 Chapter 6: Forms of the Momentum Equation: Equations of Euler and Navier-Stokes
Reynolds numbers, laminar flow is unstable if the distur-
bance is large enough. It is possible, with many precautions,
to obtain a laminar flow in a very smooth pipe up to a
Reynolds number of 40,000, although under normal
conditions the critical value of Reynolds number for a pipe
is 2,000. A disturbance superimposed on the primary
motion causes, as previously seen, a large local increase
in the convective inertia forces.
PROBLEMS
6.1 Consider successively a circular pipe and a square pipe
rotating around their own axes at an angular velocity
varying suddenly from w = 0 at time t = 0 to w = w 1
at timet = t:(w 1 small) and w = w 2 at timet = t dw 2 large).
These two pipes are successively half filled and fully filled
with liquid. Describe qualitatively the liquid motion in
the two cases where (1) the fluid is perfect, and (2) the fluid
is viscous.
6.2 Demonstrate that the velocity distribution for a flow
between two parallel planes, one of them being fixed and
the other one moving at a constant velocity U, is u = Uy/e
where e is the distance between the two planes, and under
a gradient of pressure dpjdx, the flow between two fixed
parallel planes is
dp 1 2
u = - -- (ey - y )
dx 2J1
6.3 Write a Navier-Stokes equation for an unsteady flow
between two parallel planes in which one of the planes is
fixed, while the other one is moving at a speed u(t).
Then write the Navier-Stokes equation for a two-
dimensional steady flow between two planes almost
parallel; one plane is fixed, and the other plane is moving
at constant velocity U. Do the simplifying approximations
that you think are permissible for analyzing the flow motion.
6.4 Calculate the two-dimensional velocity distribution 6g
u(z) between two parallel horizontal planes between which
there are two layers of fluid of thickness e 1 and e 2 , viscosity
J1 1 and Jlz, and density p 1 and p 2 (p 1 > p 2 ), respectively.
One plane is fixed and the upper plane moves at constant
velocity U.
6.5 Consider a two-dimensional flow between two fixed
parallel horizontal planes separated by a distance 2h.
1. Write the continuity relationship, the Navier-Stokes
equation, and the boundary condition. The flow
motion will be assumed to be in the 0 X direction
and OZ is perpendicular to the plane.
2. j being the head loss defined by dpjdx = - pgj,
calculate the velocity distribution u = fU,z) by two
successive integrations, and the total discharge per
unit of width Q = f(j,h).
3. Calculate the mean velocity u = fU,h) and express u
as a function of u, z, and e.
4. Calculate d 2 ujdz 2 and dpjdx = f(u,h).
5. Calculate the rotational coefficients ~. 1'}, (as functions
ofj, z, h.
6. Calculate the loss of energy per unit length of the
direction of the flow: pgjQ = fU,h) and the value of
j as a function of Q and h.
7. Should an obstacle be inserted between the two planes,
demonstrate that the mean motion with respect to the
vertical OZ is irrotational, i.e., (8uj8y) - (8vj8x) = 0.
(It is the Hele-Shaw analogy.) Express the potential
function as a function of p, h, and Jl.
6.6 Calculate the ratio of inertial force to viscous forces in
the case of a laminar steady uniform flow. Discuss the
statement which consists of saying that the Reynolds
number is a significant dimensionless parameter giving
the relative importance of the inertial force to viscous force.
Is the ratio of the gradient of kinetic energy to the viscous
force a more significant definition?
 Part 1 : Establishing the Basic Equations that Govern Flow Motion

70 6.7 The following dimensionless quantities are defined: 6.9 Demonstrate that the Navier-Stokes equation can
still be written:
z
x* = -
X
y* =I z* =-
L L L p d~ - ll'~jl~ = p[~ ou + 11 ou + 'ou]
dt ax oy oz
t V*(u*,v *,w *) -_ V(u,v,w)
t* = -
T
u Obtain the two other equations by circular permutation.

6.10 Establish the equation of flow for a laminar flow in an

p* = ___!!_
pU2 horizontal cylindrical pipe of radius R. Determine the
velocity distribution u(r) and the relationship between the
where L, T, U are an arbitrary typical length, time, and total discharge Q and the head-loss. The head-loss is
velocity, and F is the gravity force. Then demonstrate that defined by j = (1/pg)(opjox) where opjox is the pressure
the Navier-Stokes equations can be written in dimension- gradient along the pipe.
less form as:
6.11 Establish the equation of motion for a horizontal
steady flow in an unbounded ocean subjected to a wind
stress -r and Coriolis acceleration. It will be assumed that
the fluid is viscous, the free surface horizontal, and the pres-
sure gradient at the free surface negligible (opjox, opjoy = 0).
The wind stress -r is in the 0 Y direction, the axis 0 X and
OY are horizontal, and the axis OZ is positive downward
and two other similar equations. Explain the physical from the free surface. Draw three-dimensional diagram
significance of the parameters: of the velocity vector V as a function of depth (called
Ekman spiral).
UT
(sometimes called reduced frequency)
L
u2
(Froude number)
gL
UL
(Reynolds number)
v

6.8 Demonstrate that in a flow defined by v = w = 0 and

u = f(y,z) =f. 0, one has
 Chapter 7
71
7-1 The Definition of Mean Motion
and Mean Forces
7-1.1 Characteristics of Mean Motion
vs Actual Motion
Turbulence: In the previous chapters, theory was sometimes illus-
Mean and Fluctuating trated by examples in which the real motion was actually
turbulent, despite the fact that we were dealing only with
an ideal fluid.
Components of In these examples, it was implied but not specified
that only the average values of the velocity and the pressure
Motion were considered. In a turbulent motion, the true velocity
and pressure vary in a disorderly manner. In fact, a turbulent
motion is always unsteady, since at a given point the
velocity changes continuously in a very irregular way.
It is also nonuniform, since the velocity changes from point
to point at a given time, and rotational, since the friction
forces, proportional to V2 V, are important. These character-
istics are true as far as the actual motion is concerned.
However, by splitting the motion into mean and fluctuating
components, the average motion may often be considered
steady, uniform (in a pipe), or irrotational (over a weir).
The previously considered examples were relative to the
average values.
Now this method has to be justified and the differences
between the motion of an ideal fluid or a viscous flow,
and a mean turbulent motion have to be further considered.
This is the purpose of this chapter.

7-2.1 Validity of the Navier-Stokes

Equation for Turbulent Motion
Equalities between the inertia forces and the applied
forces on an elementary fluid particle are valid even if the
motion is turbulent. Hence, the basic Navier-Stokes
equations and continuity relationships are also theoretically
valid in the study of turbulent motion.
It has been seen that it is sometimes possible to calculate
a laminar solution where the boundary conditions are
simple. It is also possible to determine theoretically the
 Part 1 : Establishing the Basic Equations that Govern Flow Motion

72 stability of the solution, that is to determine whether a sum of the mean pressure p and a fluctuation term p' such
small disturbance will increase or be damped out by friction. that p = p + p' where
However, a fully deterministic approach is no longer
possible in the case of turbulent motion, because of the p= T
1 fT p dt and p' =
1
T
fT p' dt = 0
0 0
random nature of these turbulent fluctuations.
On the other hand, in engineering practice, it is not Hence, turbulent motion may be considered as the super-
always necessary to know the exact fine structure of the position of a mean motion and a fluctuating and disorderly
flow. Only the average values and the overall and statistical motion, random in nature, which obeys statistical laws.
effects of turbulent fluctuations have to be studied.
7-1.4 Steady and Unsteady Mean
7-1.3 Definitions of the Mean Values in a
Turbulent Flow Turbulent Flows
In a turbulent motion, as in the case of a viscous flow, The mean value is defined for intervals of time T, which
velocity and pressure have to be known as functions of the are large compared to the time scale of turbulent fluctua-
space coordinates and time. tions but small compared to the time scale of the mean
The instantaneous velocity V at a fixed point is the motion.
vectorial sum of the mean velocity V with respect to time If, for example, one considers the oscillation of water in a
(referring to the basic primary movement) and .the fluctua- tunnel between a surge tank and a reservoir, the instan-
tion velocity V' which varies rapidly with time both in taneous velocity at a fixed point may vary quickly because
intensity and direction. This can be expressed by the of the turbulence. The average velocity defined for a
relationships V = V + V' where, by definition, relatively short interval of time varies also with respect to
time, but its change is slow.
V = _!_ fT V dt The real motion is always unsteady because of turbulence
T o and in this case, the mean motion is also unsteady (Fig. 7-1).
and In the following discussion, a motion is called unsteady
only if the mean value of the velocity varies. The interval
of time, which permits a realistic definition of the mean
V' = _!_ fT V' dt = 0
motion, is relative to the frequency of turbulent fluctuations.
T o
It varies with the phenomenon to be studied. For example,
where T is a time interval. Similarly, the instantaneous it is long for the meteorologist who deals with atmospheric
components of velocity are defined as follows: motion, and it is short for the aerodynamist who deals
with the turbulence effects in the boundary layer along a
Real Mean Fluctuation wing.
velocity velocity velocity
7-1.5 Mean Forces
u ii + u'
v i5 + v' Since the real value of the inertia forces is always equal to
w w + w' the sum of the real values of the applied forces in any kind
of motion (laminar or turbulent), the mean value of the
inertia forces with respect to time is equal to the mean
and u' = v' = w' = 0. value of the applied forces with respect to time. This may
Similarly, the instantaneous pressure p IS the scalar be expressed as shown in Equation 7-1. Since
 Chapter 7: Turbulence: Mean and Fluctuating Components of Motion

v v 73

Figure 7-1
The steadiness of a turbulent flow is
defined by the mean velocity only.
Vaverage = constant Vaverage = f (t)

TURBULENT STEADY FLOW TURBULENT UNSTEADY FLOW

Local .
Convective p ressure Gravzty F . . Now each of these mean forces has to be expressed as a
. . . . rzctzon
fi
znertza + znertza + orce + fiorce + fiorce = 0 function of the mean values and fluctuating values of the
fiorces fiorce velocity and the pressure. For this purpose, one considers:

one also has: 1. The constant forces Gravity force

2. The linear forces Pressure force, linear function
Mean value ~oc~l C~nvec_tive ofp
with respect [ znertza + znertza (7-1) Local inertia force, linear
to time of force force function of V
Friction force, linear function
+ Gravity + Pressure + Friction] = 0 ofV
force force force 3. The quadratic force Convective inertia, function of
V 2 or product of two com-
Equation 7-1 is expressed mathematically, along the OX ponents of velocity: u 2 , v2 ,
axis, as (see Section 6-1.2.1) w2 , uv, uw, vw

-1 JT p(au
- + uau
- + v- au) dt
au + w-
7-2 Calculation of the Mean Forces
T o at ax ay az

= ~sT
T 0
(- o(p + pgz) + J.1'V2u) dt
ox
The mean forces are calculated as a function of the mean
values of velocities and pressure. For these calculations,
it is assumed that the order of mathematical operations
or, using the a and T notations and the rotational coefficients has no effect on the final result. In particular, integration
1J, (,and ~ (see Sections 6-1.2.2 and 6-1.3) during an interval of time T and derivatives with respect
to time or space can be interchanged.
- + -a (V
JT p (au 2
-1 - ) + 2(W1J - v() ) dt
r o at ax 2 7-2.1 The Constant Force
The gravity force depends only on the density of the
1
=+ T
JT (
0 -
a
ox pgz + ox
a (Jxx
a
+ oy Lxy
a
+ oz Lxz
)
dt elementary particle. The fluctuations of pressure are too
small to have a significant effect on the density. Hence, the
Similar equations are found for the 0 Y and OZ axes. gravity force is the same for laminar and turbulent motion.
 Part 1 : Establishing the Basic Equations that Govern Flow Motion

74 The mean value of the gravity force is equal to this constant 7-2.2.3 The viscous force has a mean value which may
gravity force. Mathematically, this may be expressed be calculated by considering, for example, one of the second-

pg =T
1 r
Jo pg dt = pg T Jo dt = pg
1 rT
order terms, such as Jl.(0 2 ujox 2 ). Averaging this term
leads also to

Since pg is a constant with respect to time, then

0 0
oz (pgz) = oz (pgz) The mean viscous force is equal to the viscous force due
to the mean velocity alone, and is mathematically expressed
in the same way as the actual motion.
7-2.2 Linear Forces All the linear forces involved in the mean motion are
7-2.2.1 The mean value of the local inertia force, may mathematically written in the same way for both mean
be obtained by considering the mean value of any of its turbulent flow and actual motion, turbulent or laminar.
components. For example,

~
p- = -
1 iT pOU
0 1
- dt = p - -
ot r
iT
o
u dt
7-2.3 The Quadratic Forces
Consider the component u = u + U 1
Squaring and
ot r o ot

averaging u leads successively to

Introduce u = u+ U 1
, where U1 is a fluctuation term, into
this equation.
u2 = ]__ rT 2 u dt = ]__ ((u 2 + 2UU 1 + U 12 ) dt = u2 + U 12
Jo T Jo
~
p- = p0- -
1fT(u + u)dt 1
T
u
at at r o This is a result of the following intermediate steps. Since
is a constant in the interval of time T:
= p ~ ]__ ( u dt + p ~ ]__ ( U1 dt
otT ) 0 otT ) 0
-1 fT u dt = u
2 2
Since T o
]__ ( u dt
T Jo
= u and u = -1
T
I iTu do
I
t= 0 Since U = 0:
1

the mean value is given by: 1

T Jr0 2uul dt = 2u T
1
Jr0 u dt =
1
0
au au
Pot =Pat Since U 1 may be positive or negative but U 12 is always
positive and its mean value is different from zero:
Hence, the mean value of the local inertia force with
respect to time is equal to the inertia force caused by the
change of value of the mean velocity alone. -
T o
liT U 12 dt = U 12

7-2.2.2 Similarly averaging the pressure forces yields, Similarly, the mean value of uv is
for example:
op
--=--
av
ox ox and the mean values of UV and U V are zero.
1 1
 Chapter 7: Turbulence: Mean and Fluctuating Components of Motion

Now, consider the mean value of any term of convective 75

inertia, such as pu(oujox). One has successively, 7-3 The Continuity Relationship
In the simple case of an incompressible fluid, the con-
au 1 IT au 1 IT a tinuity relationship is written (see Section 3-2):
~~=To~~~=~o~+~~~+~~

= -1 IT(-u- _ 0U + u -OU + u -0U') dt

OU + u- 1
1 1

T 0 ox ox ox OX
Considering each of these terms independently one obtains, This relationship, expressed as a function of the mean
since u and oujox are constant with respect to time, components of velocity and turbulent fluctuations, becomes
o a a
IT au
1
T u ox dt 0 =
au
u OX
ox (u + u') + oy (v + v') + oz (w + w') = 0
or
1IT ou'
T o u ox dt =
1 0 IT
u T ox o u' dt = 0 ou ov ow ou' ov' ow'
-+-+-+-+-+-=0
ox oy oz ox oy oz
-1 ITu,ou- dt -- -1 au- IT u,dt-- o
T 0 ox T ox 0 The averaging process, applied to oujox, gives
1 ITu' ou' 1 a IT u' 2 a u' 2 ~
T ox dt = T ox 2 dt = ox 2 = u' ox =f. 0 au au
0 0
ox ox
Introducing these values yields au'jox gives
and applied to

au (_au
pu-=p u-+ ~)
u- ou'
-=-=0
ou'
ox ox OX ox ox
Similarly, it is found that
Then the continuity relationship for the mean motion
becomes
ov ~)
8v ( u-+u-
pu-=p
OX ox ox
and so on.
Hence, the mean value of a convective inertia force with Con seq uen tl y,
respect to time is equal to the sum of the convective inertia
caused by the mean velocity and the mean convective ou' ov' ow'
-+-+-=0
inertia caused by the turbulent fluctuations. As far as the ox ay oz
mean value of the velocity alone is concerned, the convective
inertia terms have the same mathematical form as for the The mathematical form of the continuity relationship
case of a laminar motion. is the same for the mean motion as for the actual motion.
Part 1 : Establishing the Basic Equations that Govern Flow Motion
7-4 The Characteristics of the
Mean Motion of a Turbulent Flow
Insofar as the mean velocity and the mean pressure
alone are concerned, the basic momentum equation and the
continuity relationship have exactly the same mathematical
form as the corresponding equations for the actual motion.
However, other forces exist and have to be added. These
new forces are caused by the convective inertia of the
turbulent fluctuations. If these "new" forces may be
neglected, or as long as only the forces which are functions
of the mean velocity and mean pressure are dealt with,
the solutions of problems concerning turbulent motion
have the same mathematical form as the solutions given
by the Navier-Stokes equations. For example, a mean
motion which is steady and irrotational and for which the
viscous forces ~tV 2 V are neglected obeys the Bernoulli
equation:
-2
p .!':.__ + p + pgz = constant
2
as found in Section 6-1.2.5 where the velocity V and the
pressure pnow designate the average values. When applying
this equation, the assumptions must be made relative to the
mean motion; i.e., the mean motion must be steady,
irrotational, and without viscous friction (despite the
fact that the actual turbulent motion is always unsteady,
rotational, and dissipative).
In practice the fluctuations of pressure p' are very small
by comparison with the real pressure p, so that p ~ p. On
the other hand, the viscous forces ~tV 2 V caused by the mean
motion are generally small in comparison with the other
forces, in particular with the convective inertia forces
caused by the turbulent fluctuations. The viscous forces
can often be neglected except, for example, in a laminar
boundary sublayer between a smooth wall and a turbulent
boundary layer.
Now the effects of the convective fluctuating forces on the
mean motion have to be studied. Then a relationship
between the value of the mean velocity and the fluctuating
velocity has to be established. Indeed since another un-
known V'(u',v',w') has been added, another relationship
is necessary. This will be the purpose of the next chapter.
7-5 Reynolds Equations
7-5.1 Purpose of the Reynolds Equations
Expressing each force in the Navier-Stokes equation
as a function of the mean values V(u,v,w) and the fluctuating
values V'(u', v', w') and averaging, leads to the Reynolds
equation. The Reynolds equation is the form of the Newton
or momentum equation for turbulent motion.
Since each of the mean forces has been calculated in the
previous sections, it is possible to obtain directly the
Reynolds equations. To do this, the sum of these mean
forces is equated to zero. Recall that each force has the
same mathematical form as in the Navier-Stokes equation,
but it is expressed as a function of the mean values of
velocity or pressure. However, additional convective inertia
forces exist, caused by the fluctuating terms. For example,
the mean value of the quadratic inertia term, pu(au;ax), is
_au au'
I
pu ax+ pu ax
Hence, the momentum equation valid for the average
motion in the OX direction may be written directly as
Equation 7-2. (Since the calculation method is identical
in the 0 Y and OZ directions, only the momentum equation
along the OX axis is studied.)
7-5.2 Reynolds Stresses
The convective inertia caused by the fluctuating velocity
components is given in Section 7-2.3 as
---ail + at!
p ( u' ax
~)
v' ay + w Tz
 Chapter 7: Turbulence: Mean and Fluctuating Components of Motion

77
p(-au _au _au _ au ----:a;! , ou' ---:a,/)
+ u- + v- + w- + u - + v - + w - = -
a _
- (p + pgz) + Jl'V u
2_
(7-2)
at ax ay az ax ay az ax
Local Convective inertia Convective inertia Pressure and Viscous
inertia caused by the mean caused by the fluctuation gravity forces force
velocities velocities

The value of 7-5.3 Value of the Lame Components in

a Turbulent Motion
pu'(ou'
- ov'
- +ow')
ox+ ay az- The applied forces have been expressed independently
of their physical nature, as is shown in Section 6-1.3.
was shown to be zero by the continuity relationship in It is recalled, for example, that the applied forces along the
Section 7-3. Thus, this expression can be added to the 0 X axis are expressed by
convective inertia without changing the value. When these
two expressions are added and grouped in pairs, the result is: X+
o(Jxx orxy orxz)
( -+-+-ox ay az
The averaging process applied to these terms (which are
either constant, such as X, or linear) gives for the applied
forces:
This can be written as

ou' 2 ou'v' ou'w')

p(- + - + - -
OX oy az
Now, by introducing these terms (and two similar terms
obtained for the OY and OZ directions) into the general Introducing this above expression in the Reynolds
momentum equation, the Reynolds equations (Equation equation yields
7-3) are obtained.
du a - -;z
It is seen, indeed, that these Reynolds equations are p dt = X+ OX (CJxx- pu )
very similar to the Navier-Stokes equations shown
previously. The difference is in the convective inertia forces a_ -,-, a_ -,-,
caused by the turbulent fluctuations and in the fact that + oy (rxy - pu v) + oz (rxz - pu w)
the other forces are expressed as functions of the mean
value of the velocity or pressure. From this equation it is easily deduced that the fluctua-
The turbulent fluctuation forces, are called "Reynolds tion terms may be considered as external forces which
stresses.'2 are added to the other forces defined by normal forces ii
Part 1 : Establishing the Basic Equations that Govern Flow Motion

oii
p( ot
- oii - oii
+ u ox + v oy + w oz
- oii)
= -
a - + pgz) + J.1V 2ii -
ox (p
(ou' 2 ou'v'
p ox + ay + Tou'w')

(7-3)

ow ow ow ow) o (ou'w' ov'w' ow' 2)

p( at + ii OX + ii oy + w oz = - oz (p + pgz) + J.1V 2 w- p ----ax + ay + fu
'----v------' '-,----/ ' - - - - - . , - - - - - '
Local Convective Pressure and Viscous Turbulent
inertia inertia gravity forces fluctuation
forces forces

and shear stresses i. Hence, these new external forces to be their space derivatives is unaltered by a change in the
dealt with are: axes of reference. In particular:
Normal force: [ ] = (J- XX
(J XX -
--;z
pU u'v' = u'w' = v'w' = 0
_
= -p
au
+ 2f.l-- pu
--;z It is evident that isotropy introduces a great simplifi-
OX cation in the calculations. However, because of the
boundaries, the turbulence is not isotropic and the products
Shear stress: u'v', u'w' and v'w' may differ from each other. There exists
_ (au av) -puv
=-p+f.l-+-
---,, a correlation between u' and v', u' and w', and v' and w',
ay ax defined by the coefficients:

and so on. These new total external forces may also be u'v' uw v'w'
defined by a tensor of rank two similar to the first tensor (u'zv,z)lf2' (u'2w'2)1;2' (v'zw,z)l/2
defined in Section 5-5.2.
In practice the viscous forces caused by the mean These coefficients are equal to zero in the case of isotropic
velocity are very often negligible in turbulent flow in turbulence. Since the convective inertia forces caused by the
comparison with the other forces, and particularly in fluctuation terms are functions ofu' 2, v' 2, w' 2 , u'v', u'w', v'w',
comparison with the shear stresses caused by the fluctuation they may be expressed directly as functions of the coefficients
terms pu'v', pu'w', and pv'w'. of correlation which are dimensionless.

7-5.4 Correlation Coefficients and PROBLEMS

Isotropic Turbulence
By definition, in isotropic turbulence the mean value of 7.1 Express grad V 2 /2 and V x curl V in terms of u, u', v,
any function of the fluctuating velocity components and v', w, w' for a turbulent flow.
 Chapter 7: Turbulence: Mean and Fluctuating Components of Motion

7.2 Express the average ratio of dilatational and shear 79

deformation in terms of mean and fluctuating velocity
components for a turbulent flow.

7.3 Draw a line u(t) at random on graph paper and deter-

mine ii and (u' 2 ) 1 12 . On the same graph, draw another line
v(t) at random and determine ii and (v' 2 ) 112 and the value
of the correlation coefficient
u'v'

7.4 Demonstrate that the Reynolds equations can still be

written:
ou _ou _ou _ou o [p* 2 ou --;z]
ot + u ox + v oy + w oz = ox ---;; + v ox - u
o [v(ou
+ oy oy + ov)
ox - ---,
u v']

+~
oz [v(oii
oz +ow)_
ox u'w']
and two other equations which will be determined. Indicate
the advantage of this form of the Reynolds equation.

7.5 Write the Reynolds equation in the case of a mean two-

dimensional motion. Write the Reynolds equation in the
case of isotropic turbulence [u' 2 = v' 2 , u'v' = 0].

7.6 Write the Reynolds equation for a flow in a straight

circular pipe and demonstrate that the pressure is smaller
on the axis of the pipe than on the wall.
8o
Chapter 8 8-1 Some Physical Effects of Turbulent
Fluctuations
8-1.1 Velocity Distribution: Effects of
Reynolds Stresses
Turbulence Effects: 8-1.1.1 The velocity distribution depends upon the total
shearing stresses, which are a function of the viscous force
Modern Theories and Reynolds stresses:
pu'v' pu'w' pv'w'
In the case of a viscous flow, it has been shown (see
Section 6-2.1.2), that the velocity distribution over a sloped
plane is parabolic. In the case of a turbulent flow, the
velocity distribution is influenced by the Reynolds stresses
and therefore different from a parabola. The effects of the
Reynolds stresses on the velocity distribution are analyzed
qualitatively in the following section.

8-1.1.2 Consider two flui"d layers defined by the mean

motion, i.e. separated by streamlines tangential to the
vector "mean velocity" (Fig. 8-1). Let VI and v2 be the
mean velocities of these two layers in a given cross section.
The instantaneous velocity V 1 is the sum of the mean
velocity V1 and a fluctuating term V'1 , V 1 = V1 + V~.
V'1 has two components: v'1 , normal to the mean velocity,
and v~ in the V1 direction.

Figure 8-1 Change of momentum by turbulence

between two stream tubes.

..[]
Figure 8-2 The shearing stresses caused by
turbulence decreases near the
boundary, while the viscous force
increases.
Because of the normal component v'~> an amount of
fluid moving in the V1 direction at the mean velocity
V 1 penetrates from layer 1 into layer 2, and since its mean
velocity V 1 is smaller than the mean velocity V2 of layer 2,
this amount of fluid tends to slow down the speed of layer 2.
Similarly, because of the fluctuating components of the
velocity, the amounts of fluid penetrating from layer 2
into layer 1 tend to increase the velocity of layer 1. In a word
because of the turbulence, the mean velocities of the two
layers tend to become equal. It is seen that the fluctuating
velocity forces act physically as a shearing stress between
the two layers of fluid.
8-1.1.3 In a turbulent flow these shearing stresses caused
by turbulence are usually more active than the shearing
stresses caused by viscosity. Therefore, adjacent layers
tend to have similar velocity as in the case of an ideal fluid.
However, at a boundary layer, the terms of the form
pu'v' tend to zero since the velocity normal to the boundary
must vanish (Fig. 8-2). Conversely, the viscous term
Figure 8-3
Velocity distribution in a pipe.
--
--1-----+----x
IDEAL FLOW
Chapter 8: Turbulence Effects: Modern Theories
J1(o2 ujoy 2 ) increases near the boundary and becomes 81
particularly important in the case of a smooth boundary.
The mean velocity distributions in a pipe, given by
Fig. 8-3 and corresponding to different assumptions made
on the shearing stresses, illustrate these previous considera-
tions.
The quantitative study of the velocity distribution in a
turbulent flow depends upon the assumption made on the
distribution of the value of the shearing stress 1:. This is
the subject of the following sections.
8-1.2 Irrotational Motion
A turbulent motion is strongly rotational since the actual
friction forces have an important effect. However, rotational
motion occurs at random, like the turbulent fluctuations.
In the case of isotropic turbulence the mean motion
could be considered as irrotational. Where the turbulence
is nonisotropic, the mean flow is rotational, but out of the
boundary layer the turbulence is nearly isotropic in a first
approximation. Hence, a number of methods of calculation
which give the flow pattern in an ideal fluid may be success-
fully applied in a turbulent flow, as long as the boundary
layer is thin with respect to the main flow.
It is evident that such an assumption is of particular
importance in engineering practice since it permits a
knowledge of the flow pattern of the mean motion in any
convergent short structure, such as a bellmouth gallery
or a spillway (Figs. 2-12 and 2-14).
These considerations are also illustrated by Fig. 8-4.
-
-
~
--+----+---X
_.-/
LAM INAR FLOW TURBULENT FLOW
 Part 1 : Establishing the Basic Equations that Govern Flow Motion

82

IDEAL FLUID , LAMINAR FLOW:

IRROTATIONAL 1\iiOT ION Figure 8-4
ROTATIONAL MOTION
A turbulent flow may often be considered as
irrotational.

TURBULENT FLOW;
MEAN MOTION IRROTATIONAL
EXCEPT IN THE BOUNDARY LAYER

8-2 Pressure Force. Secondary Effects pressure value. For example, it is seen in elementary
hydraulics that the hydraulic jump theory is developed
8-2.1 Paradox of Du Buat by equating the variation of momentum to the external
Considering the normal forces forces, i.e., the difference of pressure forces before and after
the jump (Fig. 8-5). To be more exact, it would be necessary
[a ] = - p ou
+ 2f1 ox - ""i2
pu to add to the pressure forces the difference in Jg (pu' 2 ) dh
(h is the water depth). This term is actually negligible.
It is seen that an additional force has been added to the However, this same factor is sufficient to explain why the
pressure force. This results in an increase of the average resistance of a body moving with velocity V in calm water
 Chapter 8: Turbulence Effects: Modern Theories

,2 ,2
--+-+- pU pU

Figure 8-5
Turbulent fluctuation terms have to be added to the
pressure forces.

is different from the resistance of this same body when Finally, the Reynolds equations are reduced to :
stationary in a turbulent flow of the same mean velocity V.
This is the paradox of Du Buat. This phenomenon is caused op* o2 u ou'w'
by the difference of impulse of the turbulent convective 0= -Tx + f.l oz 2 - p ----a;-
inertia pu' 2 acting against the body in a manner similar to
op* ow' 2
pressure forces. o=- ---a; -P az
where p* = p + pgz. For a laminar flow between two
8-2.2 Turbulent Flow between Two parallel planes the Navier-Stokes equations would be
Parallel Planes written :
8-2.2.1 Consider the simple case of uniform steady
turbulent motion between two horizontal parallel planes,
as shown by Fig. 8-6. Since the mean motion is steady the
local inertia forces are zero : p(ouj ot) = 0; p(ovj ot) = 0;
p(owjot) = 0. Assuming that the mean velocity vector
is parallel to the two planes in the 0 X direction, the com-
ponents v and w along the OY and OZ axis respectively Figure 8-6 Turbulent flow between two
are zero and all the terms of the Reynolds equations where parallel planes.
those quantities appear are zero (see Section 7-5.2). Since
the mean motion is uniform, ouj ox = 0. It is then seen that
all the convective inertia terms are zero as in any case of
uniform flow.
Now consider the fluctuation terms. The derivatives of
u' 2 , v' 2 , w' 2 , u'v', u w
' ', v'w' with respect to x are zero since
the motion is uniform. The derivatives with respect to y are
also zero since the two planes are assumed to be infinite,
and the motion is two-dimensional. y
Part 1 : Establishing the Basic Equations that Govern Flow Motion
8-2.2.2 Integrating the second of the above equations
with respect to z yields
p* + pw' 2 = constant
Let the pressure at the boundaries be p6. Since w' = 0
at the boundaries, the mean pressure p* at any point of the
flow is smaller than the mean pressure at the boundary
p6 by the quantity pw' 2
8-2.2.3 If D0 is the mean velocity between the two planes,
it has been found experimentally that u' 2 !D6 < 0.01 and
v' 2 !D6 and w' 2!D6 < 0.0025. Hence the difference:
P6- p* w'z
2-
"--'---~=--
(p/2)U6 - D6
is always smaller than 0.0050 and is neglected. The pressure
distribution in a turbulent uniform flow is hydrostatic
(at least within 0.5 %).
8-2.3 Secondary Currents
Such variations of pressure caused by the fluctuation
terms explain the origin of secondary currents in straight
channels and non-circular pipes. Secondary currents take
place when nonsymmetrical effects of the turbulent shearing
stresses exist in the flow, that is, each time that the boundary
is noncircular.
These secondary currents go from the zone of high
shearing stress to the zone of lower shearing stresses as
shown by Fig. 8-7. They have a tendency to equalize the
shearing stresses at the boundary. Hence they are secondary
Figure 8-7 Secondary currents in a triangular pipe.
only in name: they partly justify the use by engineers of the
concept of hydraulic radius (Section 14-1.1). Indeed, the
definition of hydraulic radius is based on the assumption
that the shearing stress at the boundary is a constant. The
limitations of applicability of the hydraulic radius definition
must be known. A change in the secondary current pattern
in a flow has an effect on the head loss which is not negligible.
However, this effect is neglected in hydraulics because it is
not well known yet, and remains small.
8-3 Modern Theories on Turbulence
8-3.1 The Unknowns in a Turbulent Flow
In Chapter 1, it was seen that problems in hydrodynamics
consist of determining the four unknowns, u, v, w, and p.
For turbulent flow, the four unknowns are u, v, w, and p.
However, four other unknowns, u', v', w', and p' have been
introduced, theoretically requiring four other equations
(unless the fluctuation terms may be neglected). The
fluctuation values u', v', w', or some function of these values,
such as u'v', v'w', u'w', are expressed as functions ofthe mean
values u, v, and w.
It is seen that p' does not appear in the Reynolds equation
which governs the mean motion because of the linearity
of the pressure forces. Moreover, p' is usually very small
in comparison with p. p' should be taken into account
statistically only for some very special problems, such as
in the investigation of the growth of wind waves under
wind action. Although progress has been made in the
statistical theory of turbulence, only isotropic and homo-
geneous turbulence is well described by use of random
functions. However, isotropic turbulence is an idealized
case never encountered as is the abstract concept of irrota-
tionality. Further investigation into nonisotropic turbu-
lence is necessary.
8-3.2 Boussinesq Theory
In order to simplify the Reynolds equation, Boussinesq
introduced the turbulent exchange coefficient , with the
same dimensions as the coefficient of viscosity Jl. In the

case of uniform flow parallel to a plane in the 0 X direction
(u = u(y), ii = 0, w = 0), is defined by the equality
pu'v' = -s(dujdy). Then the shearing stress [r] becomes
[ r] = (f.l + s)(dujdy) instead of r = f.l(dujdy). [ r] is given
by a linear relationship. From this relationship it may be
seen that the fluctuation term would act similar to the
viscous term. They are simply added linearly. However,
they are of a different order of magnitude, that is s ~ f.1
and [r] ~ s(dujdy).
Such a relationship gives a velocity distribution similar
to that obtained in a laminar flow from the Navier-Stokes
equations. Since it would be necessary to consider that s
varies with respect to space, the Boussinesq theory is a
failure. However, in some cases relative to the motion of
atmospheric layer, s is approximately a constant and the
Boussinesq assumption is applied to obtain a result at a
first order of approximation.
8-3.3 Prandtl Theory for Mixing Length
The mixing length theory has been introduced by
Prandtl by analogy with the mean free path as it is defined
in kinetic theory of gases. It is the momentum transfer
theory.
Consider the flow u = u(y), ii = 0, w = 0 parallel to
the OX axis (Fig. 8-8). The mean velocities are u and
u + du at two points on a perpendicular to the boundary
defined by y = 0.
Figure 8-8 Flow parallel to the OX axis.
u du
-u
Chapter 8: Turbulence Effects: Modern Theories
According to Prandtl, it is assumed that the fluctuation
terms u' and v' are proportional to the difference in
velocity du which is equal to: du = (dujdy) dy, such that
u'v' = -l 2 (dujdy) 2 or Iu'l and Iv'l ~ l(dujdy).l is the "mixing
length." l may be physically considered as the length which
may be traversed by a lump of fluid perpendicular to the
mean velocity vector u. It is evident that according to this
definition, l is equal to zero at the boundary since a lump
of fluid cannot pass through the boundary.
On the other hand, u'v' always has the opposite sign of
dujdy. If one considers a lump of fluid moving from the
boundary to the middle of the flow, v' > 0 and it is moving
from a layer where u is smaller to a layer where u is larger.
It causes a slowing down of the motion, hence u' < 0.
Conversely, if one considers a lump of fluid moving toward
the boundary, v' is negative while u' is positive. Since
u'v' is always negative, the shearing stresses caused by
turbulence, r = - pu'v', are positive, as is dujdy.
If a velocity distribution has been considered such that
dujdy is negative, it will be similarly found that - pu'v' is
always negative. Consequently, in any case, [r] has the
same sign as dujdy, which is emphasized by writing:
[T] = plzl du ldu
dy dy
In the general case,
_ du
[ T ] -f.l-+p du I-
12 1 - du
dy dy dy
This function may be linearized with the help of the
Boussinesq coefficients:
Despite a more complex mathematical form, the main
advantage of the Prandtl theory over the Boussinesq theory
is that it is easier to assume the value of l than the value of s.
 Part 1: Establishing the Basic Equations that Govern Flow Motion

86 wall as it is explained in elementary hydraulics. Both cases

are theoretically approximate, but give results close to
8-3.4 Taylor's Vorticity Transport
factual measurements.
Theory
Instead of considering the change of momentum from 8-3.6 Von Karman's Similarity
one layer to another as Prandtl did, Taylor considered Hypothesis
the change of moment of momentum. This theory sometimes Von Karmim tried to find a value for l independent of
gives the same result. For example, the velocity distributions the kind of flow, according to two similarity assumptions:
in a two-dimensional jet given by both theories are the same.
1. The turbulence mechanism is independent of viscosity
However, when the mixing length is a function of the normal
(except near a smooth boundary).
distance from the boundary, different results are obtained.
2. The turbulent fluctuations are statistically the same
at any point but change only in time and length
8-3.5 Value of the Mixing Length
scales.
Now the value of the mixing length has to be determined.
Various formulas are proposed. The first kind of formula From this assumption Von Karman found that
for the mixing length is purely empirical and valid only for
special cases. Some examples are: [r] = pF ldiildii
dy dy
1. At the boundary of a jet l is proportional to the as Prandtl, and
distance from the orifice.
2. Against the wall of a pipe lis assumed to be propor-
tional to the distance y from the boundary: I = ky
where k is a constant. This means physically that the
Hence
amplitude of a turbulent fluid lump is zero at the
boundary and increases linearly with the distance
from the boundary.
Introducing this value in the Prandtl formulas yields where k is a universal constant, experimentally found to be
equal to 0.4.

8-3.7 Other Theories

Other theories were proposed to improve these semi-
If [ r] is considered as a constant, a "universal velocity
empirical formulas. Particularly, in order to avoid E = 0
distribution" is obtained by integrating with respect to y:
when dii/dy = 0, for example, in the middle of a pipe,

k In y + constant )
[ r ]) /2(1
Prandtl proposed:
p
1
ii = (

[r] has been considered as a constant by Prandtl in the

theory of the boundary layer along a flat plate. The values But it is difficult to know the best value of 1'.
of the constants are determined by experiment. In a pipe [ r] These various theories, and particularly the Prandtl and
is considered as a linear function of the distance from the Von Karman theories, have been very successfully applied
 Chapter 8: Turbulence Effects: Modern Theories

in a number of practical cases (wall, pipe, etc.). However, In the case of a smooth boundary, the term diijdy is no 87
they do not seem so successful when the flow is not uniform longer negligible in the boundary layer. Hence the head
(bend, divergent, etc.). It is to be born in mind that the loss is a complex intermediate function of ii, that is, a
phenomenon of turbulence is random, and therefore non- complex function of V.
deterministic. A solution to the problem of mathematical In the case of a laminar flow, [ r] is simply equal to
representation of such complexity is found in statistical r = Jl(dii/dy) and the head loss is a linearfunction of ii(ii = u),
mechanics, as introduced by Taylor, Von Karman, and hence a linear function of V.
Kampe de Feriet. This subject is beyond the scope of this
book. 8-4.2 Work Done by Turbulent Forces:
Dissipation Function
8-4 Some Considerations on the Loss It is evident that because of the turbulence the loss of
of Energy in a Uniform Flow energy in a turbulent flow is generally much greater than
in a laminar flow.
8-4.1 A Review of Elementary The mean value of the viscous forces per unit of volume
Hydraulics has been found to be
It has been seen in elementary hydraulics that the head
loss in a uniform flow is: J1'V 2 V = J1'V 2 V + 11'V 2 V'
where
1. Proportional to the mean value through a cross
section of the velocity V when the flow is laminar. J1'V 2 V' = J1'V 2 V' = 0
2. Proportional to its square value V 2 when the flow is The term J1'V 2 V is small by comparison to the kinematic
turbulent and the boundary is rough. forces caused by turbulent fluctuations. The mean value
3. A complex intermediate function of V (Vn) when the of the viscous forces caused by these turbulent fluctuations
flow is turbulent and the boundary is smooth is zero.
(1 < n < 2). If instead of considering the mean forces, one considers
The above result may be partly explained by the following the mean value of the work done by these forces, quite a
considerations. A part of the kinetic energy of the primary different result is obtained. Consider, for example, the
(or mean) motion of a turbulent flow is continuously mean shearing force Jl(au;ax). This force is equal to Jl(aujax)
absorbed to provide energy for the turbulent fluctuations. smce
The kinetic energy of these turbulent fluctuations is a
quadratic function of the fluctuating velocities. Since all
these fluctuations are finally absorbed by friction, the loss
Jle~) = JlG:) = o.
of energy in a turbulent flow is a quadratic function of the The work done by the force Jl(aujay) in a unit of time is
fluctuating velocities. (see Section 5-5.4)
On the other hand, the fluctuating velocities are roughly
linear functions of the mean velocities. It has been seen that JlG~Y dx dy dz
and by unit of volume: Jl(au;axf. The mean value of this
[r] = J1 dii _ pu'v' = J1 dii + p/ 2 dii Idii
1

dy dy dy dy work with respect. to time is, successively,

As long as Jl(diijdy) is negligible, u'v' is of order ii 2 and the _ jl 2 2

W- -)
(au _ }(au
- 1-+ au')
-
head loss is proportional to ii 2 , that is, proportional to V 2 ax ax ax
 Part 1 : Establishing the Basic Equations that Govern Flow Motion

88 The double product is zero: Explain the limitation of these assumptions. What is the
criterion for the pressure distribution on which such
2 au au' = 2 au !___ _!_ IT u' dt = 0 assumptions are based?
ax ax ax ax T 0
so that one finally obtains: 8.2 It has been found experimentally that

2+ J(au') 2
( au)
W=J1-ax l-
ax
where D is the distance between two parallel planes. Give
The second term is always positive, hence its mean value the expressions for [ r] and Prandtl's mixing length I as
is not zero. functions of y.
Moreover, u' is generally smaller than u, but the deriva-
tive of u' with respect to space (in this expression with 8.3 Using the Von Karman similarity rule
respect to x) is usually greater than the derivative of u with
respect to space (in this expression with respect to x). The 2 (dujdy) 4
first term (J1)(au;ax) 2 may often be neglected and W ~ r = pk (d2ujdy2)2
J1(au'jax) 2. Hence the loss of energy and the head loss are and the relationship (ap;ax) = constant along the centerline,
mainly due to the turbulent fluctuations. derive the following universal velocity distribution law
A similar result may be obtained by considering all the for a rectangular channel of width 2h
terms of the dissipation function <l> presented in Section
5-5.4 in which u, v, w are replaced by u, v, w, u', v', w'. Then
it is found that the mean value for <l> is the sum of two terms:
u_ = _u0 + -1 -
k
(r0)1/2{ [ (y)112] + (y)112}
p
In 1 - --
h h
<l>m and <l>t. <l> = <l>m + <l>t, where <l>m is a function of the
where u0 is the velocity at the centerline y = 0 and r 0 is the
mean values u, v, wonly, and <l>t, a function of u', v', w' only.
shear stress at the wall.

<l>m =
2
ax + ... + (au
J1[ 2 ( au) av) 2+ ...
ay + ~ J 8.4 It will be assumed that the velocity distribution in a
cylindrical pipe of radius R is given by the one-seventh

<l>t = J1[ 2 ( aa:') + ... + (au'

2
ay + av')
2
ax + ... J power law, i.e.,

<l>t is the part of energy which is absorbed by friction because

of the turbulent fluctuations.
PROBLEMS
8.1 Explain why [ r] is considered as a constant along a
perpendicular to the wall in the boundary layer theory
and varies linearly with distance from the wall in the case
of a uniform flow in a pipe or between two parallel planes.
where U is the maximum velocity on the centerline. Then
give an expression for the Prandtl's mixing length as a
function of r.
8.5 Determine the expression of the dissipation function
due to turbulent fluctuation as a function of au'jay only,
in the case of isotropic turbulence.
Chapter 9 9-1 Average Motion in a Porous
Medium
9-1.1 Basic Phenomenology

Flow in a 9-1.1.1 The basic laws to be applied to a flow in a porous

medium are again the continuity relationship and the
Porous Medium: momentum equation. The momentum principle, expressed
by the Navier-Stokes equations, is also theoretically
valid for this kind of motion. However, because of the
Law ofDarcy complexity of the boundary conditions (since V = 0 at the
surface of every grain of the porous medium), this equation
is no longer useful in this form. Some approximations and
transformations must be performed.

9-1.1.2 First of all, the grains are assumed to be distrib-

uted at random. The flow obeys statistical laws. (The case
of nonisotropic porous medium, such as varved clays,
necessitates the consideration of a coefficient of permeability
which varies with direction.) Hence, instead of dealing with
the real values of velocity and pressure, varying in a very
complex manner, only the mean values need to be con-
sidered.
It is evident that such a method considerably simplifies
the boundary conditions since these conditions have to be
expressed only to the boundaries of the mean flow, i.e.,
the limits of the porous medium and the free surface.

9-1.1.3 It is known that in a laminar flow the mixing

process is very slow since it is caused only by molecular
agitation, while in a turbulent flow it is rapid since it is
caused by the turbulence fluctuations. In a laminar flow
through a porous medium, because of the random nature
of the particle distribution, it may be observed that dye
diffuses quickly although the flow is laminar (Fig. 9-1). For
example, the concentration curve obtained by injecting
a colored fluid within a porous medium is given by a
Gaussian-shaped distribution.
The angle of the cone of diffusion is a function of the
characteristics of the porous medium and is approximately
Part I : Establishing the Basic Equations that Govern Flow Motion

A, V = Q/ A, independent of the void coefficient. Now,

if the motion is referred to a three-axis system of coordinates
OX, OY, OZ (see Fig. 9-2), the three real velocity com-
ponents, u, v, w are different from zero. Their mean values
in the porous medium are, respectively,

ii = _!_I
VO
Iff
vol
u d(vol) = _!_
A
If
A
u dA = V

Iff If
CONCENTRA TION CURVES

v = _!_I
VO vol
v d(vo!) = _!._
A A
v dA = 0

Iff If
Figure 9-1 Diffusion through a porous medium.
w= _!_I
VO vol
w d(vol) = _!_
A A
w dA = 0
6. This angle increases when the flow becomes turbulent;
it is then a function of the Reynolds number as defined where "vol" is the total volume of the porous medium. The
in Section 9-3.1. mean values with respect to space are written with two bars
instead of one to be differentiated from the mean value with
9-1.2 Definition of the Mean Motion respect to time {V) as used in studies of turbulent motions.
9-1.2.1 The simplest way of defining the mean velocity
consists of considering a unidimensional porous medium 9-1.2.2 For a more complicated pattern of the mean
as shown by Fig. 9-2. The mean velocity or "specific motion, where a variation of the mean value of the velocity
velocity" is the ratio of the discharge Q to the total area with respect to space also exists, the mean value of the

Figure 9-2
M ean uniform flow through porous medium.

y

velocity vector has to be defined in an elementary volume
dvol = dx dy dz of porous medium as follows:
V= --
1 -Iff
dx d y dz ~x ~Y ~z
V dx dy dz
and along three coordinate axes:
[Willv dxdydzffLx~y~z [u]v dx dy dz
~ 1
= flow.
W
iJ = 0 in the case of a mean two-dimensional flow.
Such an elementary volume of porous medium must
theoretically be large enough for the averaging process to be
valid. Hence, dx dy dz must be large enough to contain
a number of grains distributed at random.
On the other hand, dx dy dz must theoretically be small
enough to be considered as infinitely small dx dy dz in
order to apply the methods of differential calculus.
In other words, dx dy dz has to be large enough for the
averaging process to be valid, but small enough to be
considered as infinitely small in the mean motion. For both
these opposing conditions to be satisfied the gradient of
real velocity has to be much greater than the gradient of
the mean velocity. This may be physically translated as:
a large flow pattern through relatively small grains or
pebbles. A flow pattern around a few large rocks does not
obey statistical laws. Therefore it cannot be represented
realistically by a mean motion.
9-1.2.3 Similarly the mean pressure is defined by:
p- = 1
dx dy dz Iff
~x~y~z
p dx dy dz
The variations of p around pare mainly caused by the curva-
ture of the paths around the grains. Such variations are
proportional to the convective inertia, i.e., proportional
to the square of the velocity, which is usually negligible.
9-1.2.4 Considering the mean velocities with respect to
space passing across the plane sides of a cube defining an
Chapter 9: Flow in a Porous Medium: Law of Darcy
elementary volume of a porous medium, it is found by a
demonstration similar to that given in Chapter 3 that the
continuity relationship is:
au av
ax+ ay +
aw
az = 0
That is, the continuity relationship for the mean motion
has the same mathematical form as for other kinds of
9-1.2.5 The boundary conditions are expressed as a
function of the mean velocity at the boundary of the porous
medium instead of being expressed as a function of the
real velocity at the boundary of each grain.
9-1.3 Analogies between Turbulent Flow
and Flow through a Porous Medium
9-1.3.1 Interesting theoretical analogies may be made
between the methods of studying turbulent flow and flow
through a porous medium. In both cases the mean velocity
and mean pressure are dealt with because of the random
nature of the flows. In the case of turbulence the mean
values are defined at a given point with respect to time,
while in the case of flow through a porous medium the
mean values are defined with respect to space (see Section
7-1.3).
v =T-1 JTo v dt V= -
vol
1 Iff
vol
V d(vol)
The time T has to be long enough for the averaging
process to be valid, but short enough to take account of
whether the mean motion is steady or unsteady (see Section
7-1.4). The elementary volume dx dy dz must obey similar
considerations with respect to space as has been discussed
in Section 9-1.2.2.
The fluctuation terms u', v', w', p' found in the studies
of turbulence exist also with respect to space in the studies
of flow through a porous medium, and their mean value
is also zero by definition. If one defines V = V+ V', then
t1
X
1 Iff
Y
t1
Z ~x~y~z
V'(or p') dx dy dz = 0
 Part l: Establishing the Basic Equations that Govern Flow Motion

92 The momentum equation of the mean motion in a porous with respect to space as was done for a turbulent flow
medium is obtained by averaging each of the forces with with respect to time.
respect to space, as has been done with respect to time in the Since the sum of the real value of the different forces
study of turbulent flow. This will be the subject of Section involved is always zero,
9-2. Both turbulent flow and flow through a porous medium
Inertia Gravity Pressure Friction
are strongly rotational as far as the real motion is concerned.
force + force + force + force = 0
However, their mean motions may be irrotational (see
Sections 8-1.2 and 9-2.2.2). An isotropic turbulent flow may The sum of the forces' mean values with respect to space is
by considered analogous to flow through an isotropic also equal to zero
medium. Turbulent flow through a porous medium is
Mean
studied by considering the mean values with respect to
value with [Inertia Gravity Pressure Friction]
both space and time such that
respect to force
+ force + force + force =0

IT ~X ~Y1 ~Z Iff
space
=
1
V=- v~~~~
T 0 &x&y&z
9-2.1.3 For a first approximation, the inertia forces are
neglected. The local inertia is neglected because the varia-
tion of the ground water table is usually very slow. From
9-2 Law of Darcy this point of view, unsteady motions through a porous
medium are usually studied as a succession of steady
9-2.1 Approximations motions (see Section 4-5.1). However, some special problems
9-2.1.1 First of all, it must be noted that for certain require the consideration of the local inertia, for example,
flows with a free surface through a very tine porous medium, perviousness of a rocktill breakwater to periodical gravity
the capillarity forces could have an appreciable effect on waves.
the flow pattern and the discharge. The convective inertia is also neglected. Since the velocity
For example, the rise of the free surface in an earth dam is usually very small, the square of the velocity and terms
of grain size near 0.1 mm is about 1 ft (30 em) (Fig. 9-3). which are functions of the square of the velocity (such as
the convective inertia forces) are of a second order of
9-2.1.2 Insofar as these capillarity effects may be magnitude in comparison with other terms (see Section
neglected, the momentum equation expressed as a function 4-5.2.1 ). The range of validity of such an assumption is
of the mean values is given by the same averaging operation studied further.

Figure 9-3
Capillarity effects in a flow through porous medium.
 Chapter 9: Flow in a Porous Medium: Law of Darcy

permeability of the porous medium to the fluid. Hence, 93

one has finally,
9-2.2 Mean Forces
9-2.2.1 Finally the momentum equation is reduced a (~ ~~*

to an equality of applied forces: u~ = K - p + pgz ) = K -

op-
ax ax

~ ~1 ~
X Y z
Iff ( dxdydz
a(p- +
- - apgz)
--
X
+ J.i'V 2 u) dx dy dz = 0
_ a _
v = K ay (p + pgz) = Kay
ar*

Two similar equations may be written along the two a a~*

other axes OY and OZ. These three equations are, in w= K az (p + pgz) = K :z

vector form:

~ X
1~ Iff
Y Z dxdydz
[-grad (p + pgz) + J.1V' V] dx dy dz
2
or vectorially:

V = K grad (p + pgz) = K grad p*

=0 This expression defines the" Law of Darcy."
In ~lculating these equations as functions of the mean This law states that the mean velocity of the fluid flowing
value V and p, it is to be noted that they include: through a porous medium is directly proportional to the
pressure gradient acting on the fluid. P* = p + pgz is the
One constant force: Gravity force piezometric head. The simplification of the friction term
Two linear forces: Pressure force and viscous force is of an empirical nature and it seems difficult to justify
Following the same process of integration as that used such a law rigorously. It would be necessary to go through
in turbulent motion (see Section 7-2), it is found that the calculation for a flow as shown in Fig. 9-1. On the other
hand, it would seem reasonable to think that the Navier-

~ X
1~ Iff . -
Y Z dxdydz
grad (p + pgz) dx dy dz
Stokes equations are no longer valid, since from a micro-
scopic point of view, a flow passing through the very fine
channels of a porous medium, like porous china, would
=-grad (p + pgz) probably require a study based on molecular agitation. This
subject is relevant to the kinetic theory of liquids.
Now consider the viscous forces. Since they are linear, In a nonisotropic porous medium, K has different values
it is reasonable _!o expect that they are proportional to the -Kx, Ky, Kz-along the three components axes OX, OY,
mean velocity V as long as there are no quadratic effects and OZ, respectively.
caused by the convective inertia and tur~ulence. Hence
it is written that they are proportional to J.lV such that
9-2.2.2 It is important to note that such a mean motion

~X ~Y ~z
1 Iff dx dy dz
J.lV' 2 V dx dy dz
k
V
= -J.l=
defined by the law of Darcy is always irrotational.
Introducing the value u, v, wgiven above, it is easy to
verify that au;ay - av;ox = 0, since
where k is the permeability of the porous medium, k is an
empirical function of the void coefficient and grain size.
K = k/ii is the hydraulic conductivity which measures the
!___ [~
ay 1-i ox
op*] _!___ax [~1-i op*]
ay
=0
 Part 1 : Establishing the Basic Equations that Govern Flow Motion
94 z
AT TH E FRE E SURFACE
{ L'. zl = L'.z2 .. . .
L'.1 = L'.2 ... .
Similar demonstrations may be done for the two other
conditions given in. Section 2-6.2. However, if the flow
through the porous medium is turbulent, it is necessarily
rotational and hence cannot have its mean motion defined
by a potential function.
9-2.2.3 The velocity potential function for a laminar
flow through a porous medium is usually defined_ by
u = - K(o</>/ox), iJ = - K(o j 8y), w= - K(o</>/oz), or V =
- K grad . Substituting these values into the Darcy
equations gives
::1 ..1,
u'f' 0"
= - K - (p + pgz)
~ ~
u=- K -
ax ax
and two similar equations for v and w. From these, it
~easy to see that is equal to the piezometric head:
p* = (p + pgz) = . Sometimes is also defined by:
[(pjpg) + z] = p*jpg and u = - Kpg(a;ax), etc. Figure 9-4
illustrates the value and the physical meaning of corre-
sponding to such a definition. The velocity potential
function is a constant along the sides of the dike and
decreases linearly with z at the free surface inside the dike.
Hence a constant Llz corresponds to a constant value for 11.
This property will be used in Section 11-6.3 for devising
a graphical method of investigation.
Figure 9-4
Value of the potential function for a free surface
flow through porous medium.
9-3 Range of Validity of the Law of
Darcy
9-3.1 Effect of Reynolds Number
9-3.1.1 The permeability k is a function of the porous
medium characteristics and the Reynolds number. The
Reynolds n~mber of a f!ow through a porous medium is
defined by V6/v where V is the mean or specific velocity
as previously defined, and (j is the diameter of a grain.
The diameter of a grain is easily known while the" diameter
of the channels," as used for pipe, would be difficult to
define. This process assumes that there is a simple linear
relationship between these "channel diameters" and the
grain size. However, in a porous medium made of a large
grain size distribution, the small particles have a tendency
to reduce the size of the "channel." The "channels" have
the same order of magnitude as the smallest particles.
Hence, it is more exact in this case to define the Reynolds
number with the help of the smaller grain sizes. The
"characteristic diameter" (jc may be considered empirically
as the average size corresponding to the lowest 10 percent
limit.
9-3.1.2 Although the velocity in a porous medium is
very small, the derivative of velocity with respect to space
is large. It is easy to recognize this since the actual paths
 Chapter 9: Flow in a Porous Medium: Law of Darcy

in a porous medium are strongly curved. Hence the con- 95

vective inertia has an appreciable influence on the motion
when the Reynolds number is greater than one even before
the appearance of turbulence. Since the convective inertia
is quadratic, the following law of Forchheimer is more
accurate than the Darcy's law.
grad p* = av +bY lVI"
where n lies between 0 and 1.
9-3.1.3 At larger Reynolds numbers (Re > 100), the
flow becomes turbulent. The above Forchheimer equation
may still be applied but the values of the coefficients a and b
are cha_!lged. At very large Reynolds numbers, the linear
term aV becomes negligible, and the coefficient n tends
to the value 1:
grad p* bVIVI
=
9-3.2 Permeability Coefficient
The value of the permeability coefficient k is given by
dimensional analysis and experimental results. In the
general case, it is found to be a function of the Reynolds
number V6c/v, void coefficient 8, and the Froude number
V2 jg6. Many functions have been proposed, too numerous
to be compared and analyzed in this book. Some of them
are more or less theoretically justified.
For a first order of approximation, the following equation
may be used for both laminar and turbulent flow:
I'!H Cx{1 - 8) 2 V2
I'!L 83 2g()c
where pg I'!H/ I'!L is the gradient of pressure ( = p V/k)
and ex is the drag coefficient of a rough sphere for the

Then the coefficient b for the same void coefficient and the
same grain size distribution is a function of the roughness
of the rocks. A similar phenomenon has been found in
elementary hydraulics in the case of turbulent flow in a
rough pipe (see Section 13-4.3).
same value of the Reynolds number (Fig. 9-5) (see Section
13-5.1). This empirical law has been established experi-
mentally for a range of Reynolds numbers between 10 2 and
10 5 and for small variations of the void coefficient 8, such as
encountered in sand or quarry run used in rubble mound

'
"" ~

Figure 9-5
Drag of a rough sphere vs 10
""' "'
--
Reynolds number SMOOTH SPHERE-,
~ ROUGH SPHERE l

.............. l
..
'
,___ j
-2 -I 2 3 4 5 6
10 10 10 10 10 10 10 10
V8c
-v-
Part 1 : Establishing the Basic Equations that Govern Flow Motion

breakwaters and rockfill cofferdams. For a laminar motion For flow through a porous medium the mean motion
with a negligible convective inertia (Re < 1), ~ith respect to space is dealt with. The two unknowns are
V(u,v,w) and p.
24
ex= vbjv" In all cases (ideal fluid, laminar flow, turbulent flow,
flow through porous medium) the two unknowns-velocity
Hence the hydraulic conductivity is found to be equal to and pressure, real or mean with respect to time or with
respect to space-are obtained by the continuity relation-
K = ~ pgb;a3 ship and the momentum equation.
pl2p(l - a) 2

9-4.2 The Continuity Equations

9-4 A Comparison Between the Basic
The continuity relationship has the same mathematical
Equations of Different Kinds of Flow
form for four kinds of motion. It is always expressed as a
Motion function of the real velocity V(u,v,w) for an ideal fluid and a
9-4.1 The Unknowns to Be Found laminar flow:
To solve a problem in hydrodynamics, there are in ou ov 0 . d" v 0
general two unknowns to be found: the velocity V(u,v,w) OX + oy + ow
OZ = ' I.e., lV =
and the pressure p as a function of space coordinates
x, y, z, and time t. and applied as such to ideal fluid and laminar viscous flows.
However, for turbulent flow, the mean motion with It is also expressed as a function of the mean velocity with
respect to time is dealt with. The two unknowns are V(u,v,w) respect to time for a turbulent flow:
and p. The fluctuations of velocity V'(u',v',w') give rise to
some convective inertia forces acting on the mean motion au av aw 0
similar to the external forces.
-
ox+oy
- + az
- =

Convective
Pressure Viscous inertia caused
Inertia gravity friction by turbulence

du av*
Perfect fluid: Eulerian equation --
p dt ax
Laminar flow: Navier-Stokes du av*
equations p dt ax
+ f-l'Vzu

Turbulent flow: Reynolds (or dii av*

pCu' au'v' au'w)
2
Boussinesq) equations p- + p,V 2 u -- +--+ -
dt ax ax ay az
Flow through porous medium: ap* f-l~
Darcy's law when Vbjv < 1 0 --
ax
+ -u
k

and as a function of the mean velocity with respect to space
for a flow through a porous medium:
an av aw
-+-+- =
0 coefficient e. One wants the discharge to obey the rule of
ax ay az
9-4.3 The Momentum Equations
The momentum equations are written (in the table
opposite) along the OX axis only for the four cases to
be considered. Since similar terms are found in these
four equations, similar methods of integration may be
used. Some of them are only valid after some approxi-
mations or some assumptions have been used to simplify
the basic equations. For example, neglecting the turbulent
fluctuation terms and the viscous term, a turbulent flow
behaves as a perfect fluid. Hence, in order to simplify the
writing and for more generality, only the notation V(u,v,w)
and p are used in the following chapters since it is under-
~ood that V and p means V and p for turbulent flow and
V and p, respectively, for flow through a porous medium.
PROBLEMS
9.1 Calculate the total flow discharge through a porous
medium of total cross section A = 1 ftl (929 cm 2 ) and length
in the direction of the flow l = 3 ft (91.4 em) as a function
of the head. The significant grain size diameter is c5 = 0.07 in
(0.3 mm) and the void coefficient is e = 0.40. One will also
make use of Fig. 9-5 for determining the head loss coefficient.
Determine the head under which the law of Darcy no longer
applies, and the head under which the turbulence appears.
Repeat the same calculation when the porous medium is
composed of two successive layers: I= 1.5 ft (45.7 em)
and c5 = 0.12 in (0.5 mm); I = 1.5 ft (45.7 em) and c5 = 0.024
in (0.1 mm); and three successive layers of length I = 1 ft
(30.5 em) each and c5 = 0.024 in (0.1 mm), 0.072 in (0.3 mm),
and 0.12 in (0.5 mm) respectively, of same void coefficient
(kinematic coefficient of viscosity v = 1.076 x 10- 5 ftl /sec
(0.01 cm 2 /sec).
9.2 Consider a flow through a porous medium with a cross
section A = 100 ft 2 and a length in the direction of the
Chapter 9: Flow in a Porous Medium: Law of Darcy
average flow L = 100ft. One wants to build a scale model 97
ofthis porous medium at a scale A = 1/10 such that a = 1 ft 2
(929 cm 2 ) and I = 10 ft (305 em), and with the same void
similitude of Froude, i.e., q = A512 Q under a head Llhmodei =
Llhprototype x ),. For this purpose the grain size of the model
6"' will be related to the grain size of the prototype c5 P by the
relationship 6"' = KA6P. Determine the value of K in the
case where H = 100 ft (3047 em) c5 = 0.024 in (1 mm),
and e = 0.40.
REFERENCES FOR PART ONE
Daugherty, R. L., and Ingersoll, A. C., Fluid Mechanics, 5th ed.
McGraw-Hill, New York, 1947.
Goldstein, S., Modern Development in Fluid Mechanics, Vol. I.
Oxford University Press, 1938.
Hunter, R., Editor, Advanced Mechanics of Fluids. Wiley, New York,
1959.
Lamb, H., Hydrodynamics, 6th ed. Dover Publications, New York,
1945.
Landau, L. D., and Lifshitz, L. D., Fluid Mechanics. Pergamon
Press, London, 1959.
Li, W. H., and Lam, S. H., Principles of Fluid Mechanics. Addison-
Wesley, Reading, Mass., 1964.
Milne-Thomson, L. M., Hydrodynamics. Macmillan, New York,
1960.
Owczarek, J. A., Introduction to Fluid Mechanics. International
Textbook, Scranton, Pennsylvania, 1968.
Planck, M., The Mechanics of Deformable Bodies. Macmillan,
London, 1932.
Prandtl, L., and Tietjens, 0. G., Fundamentals of Hydro and Aero-
mechanics. McGraw-Hill, New York, 1934.
Rouse, H., Elementary Mechanics of Fluids. Wiley, New York, 1946.
Sabersky, R. H., and Acosta, 1. A., Fluid Flow-A First Course in
Fluid Mechanics. Macmillan, New York, 1964.
Serrin, J., "Mathematical Principles of Classical Fluid Mechanics."
In Encyclopedia of Physics (S. Fligge, Ed.), Vol. 8-1. Springer,
Berlin, 1959.
Streeter, V., Editor, Handbook of Fluid Mechanics. McGraw-Hill,
New York, 1961.
Streeter, V., Fluid Dynamics. McGraw-Hill, New York, 1948.
 PART TWO

Son1e Mathen1atical
Treatn1ents of
the Basic Equations
Chapter ro
101

10-1 General Momentum

Relationships
10-1.1 Momentum Equation Along a
The Bernoulli Streamline
The laws that govern the motion of a fluid element
Equation have been established in the first part of this book. They
are given in differential forms. The purpose of this chapter
is to establish general relationships from these equations,
the first of which gives the balance of forces along a stream-
line.

10-1.1.1 Consider first the momentum equations in the

Eulerian form:
du op*
p-=--
dt ox
dv op*
p-= - -
dt oy
dw op*
p-= - -
dt oz

For the sake of simplicity in demonstrating the first relation-

ship, it is assumed that u, v, and w vary with respect to
space only, i.e., the motion is steady and the partial deriva-
tives oujot, ovjot, ow jot, are zero.
If these equations are multiplied by the differentials
dx, dy, and dz, respectively, and then summed, the result is

= -
op*
(- dx + -op* dy + -op* dz )
ox oy oz

If it is defined that u = dxjdt, v = dyjdt, and w = dzjdt,

where dx, dy, and dz are by definition the components of an
element of the streamline dS, then the differentials can be
written as dx = u dt, dy = v dt, and dz = w dt. Therefore,
 Part 2: Some Mathematical Treatments of the Basic Equations

!02 the left side of the above equation can be written in the form

p(u du + v dv + w dw) = pd
2 yz)
(vz + v2 + wz) = d(P2- 10-1.2 The Case of Irrotational Flow
In the case of an irrotational flow, one has seen (Section
The right-hand side is the total differential of p*. Hence, 2-6.3) that the velocity vector could be defined by a potential
the momentum equation becomes function rjJ such that V = grad , and
or/J a a v = a
d(p ~2 ) = -dp* u = --
ox v=-
oy w=-
oz as
Specifying that the above equation applies along a stream- Let us consider again the momentum equation in the X
line only and that p* = p + pgz, the final result is direction, and replace the velocity components u, v, w, by
their expressions in terms of rjJ (see equation 6-1):
;~(PY; + p + pgz) = o o r/J
2 (o o
2 o 82
o 82 ) op*
p oxot + p ox ox + oy oxoy + oz oxa-;
2 = ox
When integrated this becomes
vz Since
pT + p + pgz =constant

which is the same equation that was obtained for an irrota-

tional motion (see Section 6-1.2.5). This formula has been
obtained without assuming the motion to be irrotational,
as the total convective derivatives of u, v, and w, were taken
into account exactly. However, the integration of the and the same for the z component, the above equation can
equation is limited along a streamline because of the sub- be written as
stitution of the differential dx = u dt, etc.

10-1.1.2 When the local inertia terms are introduced

and the procedures used for steady motion are applied, the
following equations result. or

a ( a v2
ax pat- + p2- + p*
)
- = 0

=constant Hence, by integrating,

or
a V2

p -V + p + pgz + p Jav
at dS pat+ p2 + p* F(y,z,t)
2
=
2 = constant

Differentiating with respect to S gives Similar equations are obtained in the Y and Z directions so
that the function F is a function of timet only. It is important
-a (pV av= 0 to point out that this equation is valid everywhere in the
2
- + p + pgz ) + p -
as 2 at fluid and in any direction, particularly along a streamline.
 Chapter 10: The Bernoulli Equation

Indeed, taking the derivative with respect to S yields applies both to steady slow motion and to the case of no 103
motion. However, in the latter case the law of hydrostatics
fP<P
P at as
a (P 2V
+ as
2
+ p*
)
= 0 is physically exact, in additio"n to being mathematically
exact, since the case of no motion implies no friction.
But, since V = a;as, then this can be written as
10-2.1.2 A similar law is obtained for a uniform steady
flow of a real fluid, i.e., with friction forces. Consider the
uniform flow studied in Section 6-2.1.1. The pressure
A number of simple cases are reviewed (vide infra), with a distribution obtained was: p = Pa + pgz cos a, where Pais
physical interpretation. the atmospheric pressure.
The isobars or lines of equal pressure are inclined at an
angle a with respect to a horizontal plane. Therefore,
10-2 Simple Cases and Physical buoyancy exerted on any body in such a flow, such as on a
Interpretations rock deposit on the bottom, is exerted at an angle a with
the vertical.
10-2.1 Slow-Steady and Uniform-Steady For a number of practical cases of flow with a free surface,
Motions a is very small and cos a may be considered equal to unity.
Therefore, the pressure distribution in a free surface uniform
10-2.1.1 As the motion is steady, the local inertia terms
flow is most often hydrostatic.
are zero; because the motion is slow, the convective inertia
This hydrostatics law is considered to be accurate
terms may be neglected; (and because the fluid is perfect,
enough even for a nonuniform and nonslow motion when
the friction forces are zero). Hence, the momentum equation
the curvature of the paths is small. The calculation of a
is reduced to an equality of applied forces-pressure and
backwater curve as seen in elementary hydraulics is usually
gravity, mathematically expressed as
based on such an assumption, even though this approxima-
ap = o tion is not always specified.
ax Figures 10-1 and 10-2 illustrate two cases in which the
convective inertia has a nonnegligible influence on the
av = o
ay
Figure 10-1 Pressure distribution is greater than
o(p + pgz) = 0 that given by hydrostatic law.
oz
---=------'---

HYDROSTATIC PRESSURE
As the axes 0 X and 0 Y are horizontal, the first two
equations show that the pressure is a constant in a hori-
zontal plane. The third equation gives
p + pgz = p* =constant
It is seen that p varies linearly with the distance from the
free surface. The law of hydrostatics is recognized; that is, BUCKET
hydrostatics could be considered as a limiting case of
steady slow motion of an ideal fluid. Therefore, the same law
 Part 2: Some Mathematical Treatments of the Basic Equations
104
REAL PRESSURE
DISTRIBUTION
HYDROSTATIC
PRESSURE
Figure 10-2 Pressure distribution is smaller than
that given by hydrostatic law.
pressure distribution, and conversely, the pressure distribu-
tion has an influence on the flow pattern. The effect of flow
curvature is studied in Section 10-2.4.
10-2.2 Slow-Unsteady and Uniform-
Unsteady Motions of a Perfect Fluid
The rotational convective inertia terms are nonlinear.
Therefore, a slow motion may mathematically be considered
irrotational and V(u,v,w) may be defined by a potential
function . The quadratic term p V 2 /2 may be neglected.
Therefore the momentum equation is
8 2-5). In a steady flow, if 8Hj8S is positive, it is because of
pat + p + pgz = f(t)
which is valid in any direction throughout the fluid. The
derivative of the function f(t) with respect to space is zero.
10-2.3 Steady Irrotational Motion of a
Perfect Fluid
10-2.3.1 The momentum equations in the OX, OY, OZ
directions of steady irrotational motion of a perfect fluid
give
v2
p2 + p + pgz = constant
(see Section 6-1.2.4). Dividing by w = pg yields the total
head H, which has the dimension of a length
vz p
-+-+z=H
2g w
The total head is the sum of the velocity head V 2 j2g,
the pressure head pjw, and the elevation head z. The value
given by a piezometer, (pjw) + z, called the piezometric
head (see Fig. 10-3).
It is seen that the variation with respect to space of the
total head H in an irrotational motion is zero: grad H = 0.
Such an equation expresses the conservation of energy of
an elementary particle of fluid as a sum of its kinetic energy,
pressure energy, and potential energy. It is emphasized that
the Bernoulli equation is valid in any direction for an irrota-
tional motion, i.e., along a path as well as along a normal to
a path. It is noticed, also, that the velocity V and pressure p
refer to the local value of the velocity and do not refer to
the mean velocity and mean pressure in a cross section of a
stream tube. The cases in which Bernoulli's equation can
be, approximately applied to stream tube are discussed in
the following section.
In engineering practice, the pressure along a boundary
may be determined by this equation when the velocity
field is given, and the flow is approximately irrotational,
as encountered in short, convergent structures (see Section
the action along the path dS of an external force such as a
pump. If 8Hj8S is negative, it is either because of the action
along the path dS of an external force, such as a turbine, or
because of friction force.
The variation in H along S measures the action of
turbo machines or the head losses in the dimension of a
length.
10-2.3.2 In a rotational flow H varies from one streamline
to another streamline, whereas His the same for any stream-
line of an irrotational flow. Figure 10-4 as compared to
Fig. 10-3 illustrates this point.
 Chapter 10: The Bernoulli Equation

105
TOTAL HEAD

FREE
SURFACE
2
v
29
VELOCITY
HEAD

Figure 10-3
In an irrotationaljiow, the total head is
a constant at any point.

DATUM

10-2.4 Pressure Distribution if! a
Direction PerpendiclJlar to the
Streamlines
10-2.4.1 In an irrotational flow, the derivative of p is
known in any direction by applying the Bernoulli equation
to the velocity field.
In a rotational flow, the Bernoulli equation gives the
derivative of p along a streamline as a function of the deriva-
tive of V, but does not give any indication of the derivative of
p in a direction perpendicular to the streamlines. However,
both of these are of equal importance in engineering
practice.
It has been seen that the pressure distribution in a
uniform flow is hydrostatic. This hydrostatic law is again
valid when the path curvature is small.
10-2.4.2 Now the general case ofnonnegligible curvature
is studied. Consider an infinitely small, curved, two-
dimensional stream tube as shown by Fig. 10-5, and an
elementary mass of fluid p dR dS in this stream tube.
Because the motion is in the direction of the stream tube,
this elementary mass is in equilibrium in a direction normal
to the streamline and is under the action of its inertia and
applied forces.
Its inertia gives rise to a centrifugal force equal to
p dR dS V 21R, where R is the radius of curvature of the
streamlines. The applied forces are the difference of pressure
forces acting on the two streamlines
+ op dR) dS - p dS = op dR dS
(p
oR oR
 Part 2: Some Mathematical Treatments of the Basic Equations

I06

VELOCITY
HEADS

PIEZOMETRIC HEAD

HEADS
Figure 10-4
PRESSURE
In a rotational flow, the total head changes
from one streamline to another.

and the gravity Integrating this equation along dR permits the calcula-
tion of the pressure distribution from the velocity fields
pg dR dS cos IX
along a curved boundary, as on a bucket of a spillway, for
From the diagram it is observed that example.
If R tends to infinity, o(p*fw)foR tends to zero and
(ozfoR) dR oz p* = constant, as has been found in the case of uniform
cos IX = dR = oR flow.

When the centrifugal force is equated to the applied forces Figure 10-5 Notation.
and the resulting equation is divided by the volume dR dS + ~dR
it gives p ~ R

V2 o(p + pgz)
p-=
R oR
or
RADIUS OF
CURVATURE

that is
 Chapter 10: The Bernoulli Equation

10-2.4.3 The above demonstration does not require the vz p 107

assumption that the flow be irrotational. Hence, it is valid
Steady irrotational flow H = -2 + -= + z = constant
g (J)

for an irrotational flow as well as for a rotational flow. It

has been seen that the pressure distribution in an irrotational Unsteady irrotational 1 a V2 p
motion - - + - + -= + z = f(t)
flow is also known by the Bernoulli equation. g at 2g w
It is valid for any direction, particularly in a direction
perpendicular to the streamlines. Hence, two methods Steady rotational flow
exist for calculating the derivative of the pressure distri-
bution in a direction perpendicular to the streamlines
Steady flow (rotational
for an irrotational flow. It is evident that the same result or irrotational)
must be obtained. This could be demonstrated by combin-
ing the above formula with a condition of irrotationality. Unsteady rotational
The simplest demonstration is that the derivative of the flow
total head H = (V 2 j2g) + (p* jw) along the radius of
curvature R is zero, i.e., All of these formulas have been obtained by exact mathe-
matical integration.
aHjaR = 0
Using the relationship demonstrated in the previous
section, one has successively 10-3 Generalized Bernoulli Equation
aH _ v av a(p*fw) _ v (av v)
10-3.1 Integration to a Cross Section
aR - g aR + aR - g aR +R
10-3.1.1 When the relative curvature of streamlines is
In an irrotational flow VR = constant; V dR + R dV = 0.
small, the Bernoulli equation can be generalized to a cross-
Hence, av jaR + V /R = 0, and therefore aHjaR = 0. This
section normal to the average streamline direction. To solve
means that the constant H is the same for any streamlines
this problem first consider the streamline. By continuity,
of an irrotational flow, as previously shown.
along a streamline: V dA = constant, where dA is an
10-2.5 Resume and Noteworthy infinitesimal area normal to the streamline. Hence, the
Bernoulli equation along a streamline can still be written
Formulas
The following formulas have been found in the case of a
perfect fluid: a [( P vz
as 2 + v* ) v dA J+ at
P av v dA = o
Hydrostatics
Steady slow motion }v* = p + pgz = constant Integrating along the streamline from position 1 to posi-
tion 2
Steady uniform flow

*) *)
(real fluid). OZ at an
p + pgz cos a = constant
angle a with the
vertical
(P 2q + Pz Vz dAz- ( P 2 Vi + P1 V1 dA 1

Unsteady slow motion p-

a
at
+ p* = f(t) + J z
1
p
av
at V dA dS = 0
 Part 2: Some Mathematical Treatments of the Basic Equations

108 This equation is now integrated across a cross section of

area A and averaged with respect to time

~for J{[(P2 ~~ + P!)v2 dA2- (P ~1 + Pi)v1 dA 1

av V dA dS Jdt = 0
+ )r2 p at 1

The value of this integration will be computed in the

following sections. [Note that the product

(P ~2 + p*)v dA
is an energy flux or power.lt is seen in elementary hydraulics
that the energy flux through a cross section A is given by
the expression

[v + pP + J
Figure 10-6 Mean velocity in a cross section.
2
1
H.P. = 550 z VA
29

in the ft-lb-sec system, where H.P. is in horse power.] Finally, V' is caused by turbulent fluctuations and is a
measure of the deviation of V from its time average. It
10-3.1.2 Across a section such as BB', the velocity V can varies both in time and direction, and is such that
be written as the vector sum (see Fig. 10-6)
V = U + U' + V' V' = s: V' dt = 0

U is in the average streamline direction and is independent

of turbulent fluctuations: Across a section such as BB', when the relative curvature
of the paths is small, all velocity vectors V are nearly
u = ~ LT ~ JJv dA dt parallel so that in first approximation U' and V' are parallel
to U, and one can write U' = uU, V' = xU; so that V ~
A
U(l + u + X), where u varies with space only, whereas x
where A is the cross sectional area and U' is a measure
varies both with space and time. By definition, both have
of the deviation of the time-average direction of V from
an average value of zero. (More rigorously x should be
the mean value U. (U' = - U at the walls.)
considered as a tensor.)

U' = ~ LT(V - U) dt 10-3.1.3 Across the section the pressure p* = p + pgz will
also vary with time and space and one can write similarly
and

If
A
U'dA = 0
p = p + p' + p"
However, the variations p' and p" are neglected.
 Chapter 10: The Bernoulli Equation

10-3.1.4 Introducing the value of V as a function of U, inertia term becomes 109

the average velocity head term becomes successively
~ Jau ds
g s at
= ~ dU
g dt
This is the formula that must be used to study, for example,
surge tanks, locks, etc.

with a = 3a 2 + 3x 2 For laminar flow in a circular pipe, 10-4 Limit of Application of the Two
there is no turbulence, V' = 0, and 3x 2 = 0. In the case of Forms of the Bernoulli Equation
a paraboloid velocity distributions, as found in elementary 10-4.1 The Two Forms of the Bernoulli
hydraulics, it is found that 3a 2 = 1. For turbulent flow Equation
3x 2 ;;:; 0.05 and 3a 2 ;;:; 0.05 to 0.01 (see Section 13-4). In the case of steady flow without friction, two forms of
Finally, consider the local inertia term the Bernoulli equation are almost identical: the first one is
_!_ IT JJ Js P avat v dA dt ds v2 p
+ - + z = constant
T Jo A
-2
g pg
Simplifying as previously seen, it becomes throughout the fluid, and the other one, valid along a

f
P s
A a(u2 /2)(1 + -;;z + x2) dS
at
=
P
A(1 + ~)u au
3 at
stream tube, is

(1
u2 p
+a)-+-+ z =constant
2g pg
Strictly speaking, neither of these equations is valid
10-3.2 Practical Form of the Bernoulli
in any real case because the conditions required for estab-
Equation for a Stream Tube
lishment of the Bernoulli equation can only be approxi-
Taking account of the above correction factors, and mated. However, they are essential in many cases. In
dividing by Q = U A, the generalized form of the Bernoulli such cases, it is important to remember the following
equation for a stream tube is assumptions:
The first form of the Bernoulli equation is valid for irrota-
(p ~~ (1 + 11.2) + P2 + pgz2) tional flow, that is, in convergent flow through short
structures. V is the local velocity. In case of turbulence, V

- (p ~i (1 + 11. 1) + jj 1 + pgz 1 ) = p Laa~ (1+ ~) dS

If a 2 and x2 and hence a are neglected, dividing by pg leads
to the common form of the Bernoulli equation
is replaced by the mean local velocity with respect to time:
V. The streamlines may be curved, but V must never be
taken as the mean velocity in a cross section.
The pressure distribution is given as a function of the
local value of V and z. The pressure distribution at the walls

g1 Js 8t
is not much influenced by the vorticity in thin boundary
( u~ p2 ) (ufg + P1 ) au dS
2g + pg + z 2 - 2 pg + z 1 = layers.
The second form of the Bernoulli equation is valid for
In the case of a uniform flow in a pipe of length L, the local unidimensional flow where the motion is rotational, but the
 Part 2: Some Mathematical Treatments of the Basic Equations

IIO relative curvature of the paths must be small. U refers to

the mean velocity in the cross section, and the kinetic head
U 2 j2g must be corrected by a factor (1 + a). tx is often
neglected in practice. It is caused by turbulence and by the
variation of the velocity in a cross section. In the case of
small curvature, the pressure distribution is hydrostatic
in a cross section or more exactly, very slightly smaller in
the center of the flow than at the boundary because of the
turbulent fluctuations (see Section 8-2.2).
The pressure distribution from one cross section to
another varies as a function of U 2 , the square mean velocity
in these cross sections.
10-4.2 Venturi and Diaphragm as
Measuring Devices
In practice it is difficult to know the exact value of the
correction factor tx. Moreover, a number of assumptions
such as small curvature of the paths, negligible head loss,
etc., are not always satisfied. It must be understood,
therefore, that a venturi used as a device to measure a
discharge by a simple application of the Bernoulli equation
from pressure variations without correction factors is not
an accurate device in itself. It is for this reason that manu-
facturers must give a calibration curve obtained experi-
mentally by measuring the discharge in a calibrated tank.
An overall correction factor, which differs for each kind of
venturi, must be given as a function of the Reynolds number
at the throat.
A similar statement could be made for a diaphragm
where the upstream pressure distribution is preferably given
by considering the local value of the velocity near the
diaphragm rather than the mean value in a cross section.
The downstream pressure distribution is roughly hydro-
static, since the flow is either almost parallel or very slow
out of the vena contracta.
10-4.3 Experience of Banki
The application of the Bernoulli equation to a venturi is
well known. Despite the approximations that have been
indicated previously, it is effectively verified that when V
increases in the convergent,p* decreases as V 112 . Conversely,
Figure 10-7 Experience ofBanki.
p* increases as V 112 in the divergent. Ifthe divergent is too
rapid, the flow separates and p* is almost a constant around
the jet. This is also the case of a jet arriving in a reservoir.
Now, one can question if any change of p* also changes
V according to the Bernoulli equation. This could be
realized by an experiment, initially conducted by Banki, in
which the pressure variations are transmitted to the inside
flow of a pipe through a membrane (see Fig. 10-7). When
the pressure within the tank increases, the pressure, trans-
mitted through the membrane in the pipe, also increases;
hence V decreases, and the rubber membrane expands.
Also, when the pressure within the tank decreases, V will
increase and the rubber membrane will contract. This
paradoxical result is in accordance with the Bernoulli
equation. However, this experience is difficult to realize
because of flow separation. This fact also demonstrates the
inherent deficiencies of the Bernoulli equation applied to a
stream tube. Finally, the change in the rubber shape changes
the flow discharge in the pipe and the motion is unstable.
10-5 Definition of Head Loss
10-5.1 Steady Uniform Flow
Consider the case of steady flow in a pipe. The head loss
may be calculated by theory in a number of cases where the
flow is laminar and the cross section is of simple shape,
 Chapter 10: The Bernoulli Equation

such as circular or square. But in the case of turbulent flow,
the value of head loss cannot be obtained by theory and
must be measured by experiment.
In this case, the Bernoulli equation permits us to define
the value of the head loss !):.H between two considered
cross sections by the difference between the total heads
at these points
measurement is locally influenced by a complex flow I I I
pattern.) However, in the case of nonuniform flow, it is
more difficult because Ui/2g is different from U~/2g; in
particular, oc 1 is different from oc 2 and they are unknowns.
Even if they are considered as known, the definition of such
a head loss involves lack of accuracy as evidenced by Fig.
10-9. Moreover, it is impossible to separate the value of

[ (1 + oc 1 ) -uz
g
p*] [
1 + ___!_
pg
- (1
g
p*]
+ oc 2 ) -uz2 + _2
pg
= !):.H
head losses caused by a succession of close sudden changes.
A linear addition of the various head losses cannot be made
because of the interaction between the effects of successive
This does not present any difficulty as long as the flow is sudden changes.
uniform. In this case
10-5.3 Head Loss in a Free-Surface
!):.H = pifpg- pjjpg Flow
It is interesting to note that any change in a free surface
10-5.2 Head Loss at a Sudden Change flow gives no extra head loss. Indeed, the initial upstream
in Flow level and downstream level are always the same provided
A sudden change in a uniform pipe, such as caused by a the considered cross sections are far enough from the
diaphragm or a bend, has an effect on the flow velocity discontinuity as is illustrated by Fig. 10-10. The increase of
distribution and the head loss at a great distance down- head loss in one place is always compensated by a decrease
stream. The head loss due to this change may be obtained of head loss in another place. Hence, the head loss definition
by extrapolating the pressure lines as shown by Fig. 10-8. of a discontinuity or sudden change in a free-surface flow,
(A pressure given by a piezometer located near a dis- such as that caused by a grid, must be specified by the
continuity, a bend or an intake has no value in evaluating relative location of two cross-sectional planes between
the head loss because the flow is not uniform. The pressure which the head loss is considered.

0>
a.
......
0..

0
I
-- -- --
--
<{
Figure 10-8 w t-------t----+---.0. H = constant
-- --
I
The head loss 11H caused by a singularity or
w
sudden change in a uniform pipe 0::
::)
is well defined by extrapolation. (f)
(f)
w
0::
0..
EXPERIMENTAL POINTS

DISTANCE S
 Part 2: Some Mathematical Treatments of the Basic Equations

112
o>
o._

'
Q_

Cl ll.H =?
<(
Figure 10-9

',--+t ----
w
::r:
w The head loss caused by a sudden change
Q: in a nonuniform flow is subject to
::::l
various interpretations.
EXPE~
(J) I
(J) zw l
w wl!>l ' ........
0:: OZI
a.. ~1
enol

DIS TAN CE S

However, in case of sediment transport, the statement
may not hold true as illustrated in Figs. 10-11 and 10-12, as
the sediment tends to accumulate upstream, proportionally
to the head loss.
10-5.4 Effect of Local Inertia upon
Head Loss
By having an effect on the velocity distribution in a pipe,
and on the corresponding shear stress, local inertia has an
important effect on the value of the head loss (see Fig. 10-13).
Head loss in a given flow at a given time cannot be
considered theoretically as equal to the value of the head
loss of the steady flow, which would have the same instan-
taneous value of mean velocity.
The head losses for the same value of the mean velocity
are different for steady flow, accelerated flow, and de-
celerated flow. This may also be noticeable for a discon-
tinuity such as the bottom orifice of a surge tank, where the
flow pattern is influenced by an instability phenomenon.
However, owing to lack of experimental data, unsteady
flows are often studied with a head loss given by an empirical
law experimentally obtained in the case of steady flow. The
studies of tide in estuary and storm surge are based on
such assumptions. As in the case of steady flow, the
Bernoulli equation may be used to define the head loss in an
unsteady flow, but even more difficulties are encountered
in determining the head loss experimentally in unsteady
flow than in steady flow.

INIT IAL FREE - SURFACE

LINE

Figure 10-10
A discontinuity in a free surface flow does not
change the total value of the head loss L1z 12 .
 Chapter 10 : The Bernoulli Equation

II3

Figure 10-11
Small dam .

10-5.5 An Application to Surge Tank
Because of its importance in engineering practice, an
example of the generalized Bernoulli equation for un-
steady motion is given. The case of a surge tank in the case
where the discharge in the penstoke is zero is analyzed
(Fig. 10-14). The application of the Bernoulli equation
between points a and b gives
v;;2g is negligible as V, in the reservoir is very small.
VV 2g is also very small and is usually neglected. Moreov~r,
Pa = Pb = atmospheric pressure. The head loss term m-
cludes the head loss at the entrance of the gallery, the head
loss in the gallery ilH = LfV 2 /2gD, and the head loss due
to the bottom diaphragm of the surge tank KVJ/2g. The
head loss in the surge tank is usually negligible as is VV 2g.
(Lis the length of the gallery andfthe head loss coefficient.)
The local inertia term is usually small enough in the
reservoir and in the surge tank to be neglected. It is taken
into account in the gallery only. Hence,

=(head loss)ab +-1 Jb -av dL. ~ Jb av dL = ~ dV = ~ dQ

g a
0 t g a ot g dt gf' dt

Figure 10-12
Large dam .
 Part 2: Some Mathematical Treatments of the Basic Equations

114 10.2 Demonstrate that the following equality is valid for

steady flow:
a (V 2 p
ox - 2 + -p +gz
2(wry - v() = - - )

Two other similar relationships are obtained by circular

Figure 10-13 A comparison of velocity distributions permutations.
within a pipe.
10.3 The velocity potential function for a flow beyond a
where Q is the discharge and f' is the cross section of the sphere of radius R is
gallery. In the case of a relatively short gallery, the value for
L could be increased by a correction factor to take into
= u(~: + r) cos e
account the local inertia of the near-radial flow near the
Determine the velocity and pressure distribution around the
entrance. Finally, with Z = za- zb, the basic dynamic
sphere.
equation for studying a surge tank is

JL ) V2 L dQ
Z =(D + ka-b 2g + gf dt
In view of the results, explain the shape that a drop of rain
where ka _ b is a friction coefficient for singularities between and an air bubble in water take.
a and b. The continuity equation is Q dt = F dZ, F being
the cross section at the free surface of the surge tank. 10.4 Demonstrate that the Bernoulli equation is valid along
a streamline using the two following fully developed forms
of the convective terms of the momentum equation
PROBLEMS
ou +v -au +w-
( ox
pu -
au)
10.1 The velocity around the limit of a circular cylinder is oy oz
given by the equation V = 2U sin ewhere U is the velocity and

J
at infinity, and where the pressure is Poo. Determine the
pressure distribution around the cylinder. {:x (~2 ) + 2(ryw - (v)

Figure 10-14 Surge tank. 10.5 Two adjacent tanks have horizontal cross sections S 1
and S 2 , respectively. The difference of level between these
two tanks at time t = 0 is h 1 - h2 = h. An orifice of cross
section A is open between the two tanks at time t = 0. The
discharge coefficient of the orifice is 0.6. Give the expression
for the time T after which the level in the two tanks is the
same.

10.6 Consider four reservoirs, A, B, C, and D, connected

as shown in Fig. 10-15 (not drawn to scale) in which the level
 Chapter I 0: The Bernoulli Equation

A B c D SURGE TANK I 15

t
60 II T T
5011

l
l
4011

1" j_ lOft

I
.. ' H
I 1 T I
M N Figure 10-16
Figure 10-15
2. Give the period of oscillation of the motion in the
is maintained at gallery. (P will be neglected for this calculation.)
A: zA =60ft (18.28 m) B: z 8 =40ft (12.19 m) 3. Give the amplitude of oscillation of z in the surge tank
C: zc = 50ft (15.24 m) D: zc = 10ft (3.04 m) in the case for which the initial discharge Q, = f~ is
suddenly stopped to a zero value and to a smaller value
respectively. The pipe between A and B is 10 in. (25.40 em) in Q; = fW~. (P will again be neglected.) Explain
diameter and 3000 ft (914.4 m) long. The pipe between C qualitatively the influence of P.
and D is 12 in (30.48 em) in diameter and 6000 ft (1828.8 m)
long, and finally, the connecting pipe MN is 5500 ft 10.8 Establish the basic equations of motion for unsteady
(1676.4 m) long, M being 1000 ft (304.8 m) from reservoir flow in parallel pipes and in series.
A, and N being 2000 ft (609.6 m) from reservoir D. The
friction coefficients f of these pipes are 0.20 for the lO- 10.9 Establish the equation of motion caused by the sudden
in. and the 12-in.-diameter pipes, and 0.224 for the pipe opening of a gate for a manifold such as that shown in
between M and N. The diameter of the pipe MN is such that Fig. 10-17.
the discharge through MN is 1.2 ft 3 /sec (33,980 cm 3 fsec).
Determine the discharges between AM, MB, CN, and ND
and the diameter of the pipe MN. Figure 10-17

10.7 Consider a hydroelectric installation including a

large reservoir where the level remains practically constant,
L
a horizontal gallery length L and circular cross section f',
a tank of horizontal cross section F and a penstoke as
shown in Fig. 10-16. The head loss in the gallery is P =
0 u-'
P0 (WfW0 ) 2 , where W is the average water velocity as a
--I
t o,
.:.:---
---1- o2
- - -----11
:....
F=- __:o_ _ ___:olf
~----~
ro
GATE 1 rL.

function of time, and subscript 0 refers to steady-state
conditions. @ r--o
1. Demonstrate that the governing equation for the fI
elevation z in the surge tank is
LF d 2 z
- -2 +z+ P=O
gf' dt
 Chapter
116
II 11-1 General Considerations on the
Determination of Flow Pattern
The laws that govern the motion of an infinitely small
particle of fluid have been established in Part One (Chapters
Flow Pattern, 2 to 9). Some general relationships among velocity, pressure,
and gravity, such as that given by the Bernoulli equation,
Stream Function, were deduced by general exact integration, independent
of the boundary conditions (Chapter 10).
Potential Function The pressure p (or velocity V) may be determined from
these general relationships after insertion of the value V
(or p).
Also, the differential equations derived from the
continuity principle and momentum equation allow us to
theoretically solve directly any particular problem, that is,
to determine the velocity V (or pressure p) when the
boundary conditions are specified. These boundary con-
ditions are peculiar for each case. An example of this is
given in Section 6-2 (laminar flow on a sloped plane).
However, the boundary conditions are usually too
complicated in the majority of cases encountered in
engineering practice, so that the use of the mathematical
theory is limited to oversimplified cases. Nevertheless, a
number of practical problems closely approximate some
simple cases that can be studied mathematically.
The purpose of this chapter is to study some of these
exact mathematical methods. Moreover, a number of
approximate methods-graphic, numerical, or experi-
mental-are based on the same mathematical principles
as those explained in this chapter. The approximate methods
extend the field of application of the exact methods con-
siderably and take into account cases in which the boundary
conditions are not so simple. The graphic flow net method
is one of these approximate methods.
It is intended that the word "exact" refer to the mathe-
matical process. The physical exactness will depend upon
the limit of validity of the basic assumptions necessary to
use such methods.
It has already been indicated that the two unknowns
to be determined are the velocity V and the pressure p
 Chapter 11 : Flow Pattern, Stream Function, Potential Function
and that theoretically, both may be found directly from the
momentum equation and continuity relationship. However,
in many cases the methods under study provide a knowledge
of the relative velocity distribution from the velocity field
calculated from the continuity principle, and an assumption
such as that of irrotationality only.
To calculate the absolute value of the velocity requires
a second step. This second step is simple when the absolute
value of the velocity at one point or at one boundary is
known.
Then, in a third step, the pressure distribution is deter-
mined by application of some of the relationships between
V and p, which have been established in Chapter 10, such
as the Bernoulli equation.
This chapter deals with the problem of the determination
of the velocity field by some analytical methods of particular
importance. These analytical methods are based on the
use of two mathematical tools that allow a concise descrip-
tion of the complete flow pattern. They are the stream
function and the velocity potential function.
11-2 Stream Function
11-2.1 Definition
11-2.1.1 The stream function is a mathematical device to
describe a flow pattern concisely by its streamlines. The
stream function may be used to calculate any kind of
flow of incompressible fluid: rotational or irrotational;
steady or unsteady; two-dimensional or three-dimensional;
laminar or turbulent; slow or nonslow motion. However,
in the case of turbulent motion, the stream function is
intended to define only the mean motion with respect to
time, i.e., the mean velocity vector V. It may also be used to
define the mean motion with respect to space of a flow
through porous medium whatever the value of the Reynolds
number, i.e., for turbulent flow as well as for laminar flow.
Although the stream function may theoretically be
defined and used for three-dimensional motion, its calcula-
tion is complex and its use has been limited. Hence, in
practice the stream function is mainly used in two-dimen- I 17
sional flow; only this case is analyzed in this book.
The stream function may be defined by any one of its
characteristics and then the other characteristics may be
deduced from this chosen definition. As was done for the
velocity potential function, the stream function will be
defined first by the velocity components.
11-2.1.2 The stream function is a natural outcome of
the continuity relationship
au ov - 0
ax+ oy-
Indeed, consider a function t/J(x,y,t) = constant such that
ot/J ot/J
U=-
oy v = - ox'
From the continuity relationship given above, it follows
that
a ot/J a ot/J _ 0
ax ay- ay ax=
which shows that t/1 always satisfies the principle of con-
tinuity; in other words, the existence of t/1 implies that the
continuity relationship is satisfied and conversely the
continuity equation implies the existence of a stream
function t/1.
11-2.2 Stream Function, Streamlines,
and Discharge
11-2.2.1 Now it is shown that such a function t/J =
constant is not only the equation of one streamline but of
any streamline of the considered flow. This is performed
by a simple change of the constant value for t/1.
For this purpose, consider the streamline equation
dxfu = dyjv (see Section 1-2.2), which may be written
u dy - v dx = 0. Introducing the value of u and v as
functions of t/1 yields the equation of streamlines in terms
of stream function
ot/J ot/J
-dx + -dy = 0
ax oy
 Part 2: Some Mathematical Treatments of the Basic Equations

118 It is the total differential dt/1 (with respect to distance) of

t/J(x,y,t). Hence, the equation of any streamline expressed
as a function of t/1 is given by the equation dt/1 = 0, or in the
case of steady flow by t/J(x,y) = constant, or in the case of
unsteady flow by t/J(x,y,t 0 ) = constant.
Changing the value of the constant gives different
streamlines of the considered flow, but the function t/J(x,y)
keeps the same analytical form. It is for this reason that
t/1 is called a stream function.
Consider the flow pattern as shown by Fig. 11-1. The Figure 11-2 Uniform flow defined by a stream
discharge dQ passing through an element dn perpendicular function.
to the streamlines is
The average value of V between A and B is
ot/1 ot/J
dQ = oy dy + ox dx = dt/1 v= ~t/1 = ~t/1
~n AB
which is also the total differential of ot/J with respect to (see Fig. 11-1).
distance.
11-2.2.2 It is verified that the stream function of a uniform
It is deduced that
flow may take the form t/1 = Ay- Bx. The velocity com-
dQ dt/1 ponents are
V=-=- aljJ aljJ
dn dn U=-=A v=--=B
ay ax
Therefore, the total discharge between two streamlines and V = (A 2 + B 2)112. V does not depend upon x and y;
t/1 1and t/1 2 is given by their difference: hence, the flow is uniform.
The streamlines are defined by the equation:
~Q = t/12 - t/11
Ay- Bx = K
They are straight lines of slope yjx = B/A and are obtained
Figure 11-1 Discharge in terms of stream by giving K various constant values as shown in Fig. 11-2.
function-notation. The discharge between two streamlines is given by the
difference between the corresponding values of the
y constant K.

11-2.3 Stream Function and Rotation

The rate of rotation is 2( = (oujay) - (av;ax) (see
Chapter 2). Expressed as a function of ljJ, the rate of rotation
becomes successively

2( = j_ aljJ - j_ (- atjJ) = o2t/J + azljJ

oy ay ax ox ay 2 ox 2
 Chapter 11 : Flow Pattern, Stream Function, Potential Function

that is, 2( = V2 t/J. Hence, an irrotational motion for which
( = 0 is defined by a stream function t/J, which is a solution
of the Laplace equation V2 t/J = 0. In other words, V2 t/J = 0
defines an irrotational motion which satisfies the continuity
principle. It may be easily verified that the example of
uniform flow defined by a stream function given in Section
11-2.2.2 is irrotational.
function permits the transformation of a system of two 119
equations with two unknowns u and v into one equation
of higher order with only one unknown.
When boundary conditions are introduced, this equation
gives the theoretical value of t/1 after successive integrations
from which u and v are afterward obtained by simple
differentiation.

11-2.4 General Remarks on the Use of

the Stream Function 11-3 Velocity Potential Function
The stream function may be used to calculate the flow 11-3.1 Use
pattern from the basic equations-continuity relationship The velocity potential function has been defined in
and momentum equation-by introducing the values Section 2-4.2. It should be recalled that the velocity potential
of u and v as a function of t/J. In this case, it must be noted function is defined as a function of (x,y,z,t) such that when
that the problem exists in finding only one unknown (t/1) differentiated with respect to space in any direction, it yields
instead of two (u and v), but from its own definition, the the velocity in that direction. For example, for one direction
order of the basic differential equation increases by one S the velocity in that direction Vsis such that Vs = (a;as).
degree. It is interesting to note the following parallel: The
For example, consider the equations which are used to velocity potential function is a natural mathematical
study the boundary layer theory as they have been estab- outcome from the assumption that the motion is irrotational
lished in Sections 4-5.2.4 and 5-4.2. (aujay) - (av;ax) = 0 in the same way that the stream
au+ av = 0 function is a natural mathematical outcome from the
Continuity: continuity relationship (au;ax) + (avjay) = 0.
ax ay
As with the stream function, the velocity potential
au au a1u function is a mathematical device to describe a flow pattern
Momentum: u ax + v ay = v ay1 concisely.
The velocity potential function may be used for any
The first equation allows definition of t/1 as
kind of irrotational flow: steady or unsteady; two-dimen-
u =-
at/f and v = -at/f
-
sional or three-dimensional. It may be used to study
ay ax turbulent motion, provided the velocity potential function
refers to the mean motion with respect to time.
Then, introducing these values in the momentum equation It may also be used to study a flow through porous
yields medium provided it refers to the mean motion with respect
atjl a1 tjl atjl a1 tjl a3 tjl to space, and that the Reynolds number is smaller than
------=V-
ay axay ax ay1 ay 3 1 (see Section 9-3.1). However, with the exception of this
last case, it may be used only when friction effects are
This involves the calculation of only one unknown t/J, negligible, and in short convergent structures. When used
but the equation is now of the third order rather than second for the divergent part of a flow, it must be realized that
order as it was when the motion was expressed by the convective inertia forces often cause separation and wakes,
two velocity components of u and v. Briefly, the stream and that the velocity potential function has a limit of
 Part 2: Some Mathematical Treatments of the Basic Equations

120 applicability. If the surface of separation of wakes is known,

the flow from stagnant zones may also be defined by a
11-3.3 Velocity Potential Function and
potential function provided the friction effects are negligible,
Continuity
as they are in the convergent part of the flow.
11-3.3.1 It is to be recalled that introducing in the
11-3.2 Equipotential Lines and continuity relationship (see Chapter 3-3.4)
Equipotential Surfaces ou + ov +ow= 0
By definition an equipotential line in a two-dimensional ox oy oz
motion and an equipotential surface in a three-dimensional leads to
motion are defined by the fact that keeps a constant value o2 o2 o2 2
at any point of this line or of this surface ox2 + oy2 + oz2 = 0 or v = 0
1/J(x,y,z) = constant = K
Similarly, introducing t/J in the equation stating that the
or flow is irrotationalleads to V 2 t/J = 0.
(x,y,z,t 0 ) = constant = K Hence, a two-dimensional irrotational flow may be
found as a solution of either
That is,
o o o
- dx +- dy +- dz = 0 or d = 0 The following table summarizes the previous con-
ox oy oz
siderations:
Changing the value of the constant K gives various equi-
potential lines or surfaces in the same way that various Continuity I rrotationalit y
streamlines were obtained when this operation was
performed with the stream function (t/1 = K). Definition of 1/1
Expressed as V2 r/> = 0
>< Definition of rj>
Expressed as V 2 1/J = 0
In contrast, the velocity vector and the streamlines are
always perpendicular to the equipotential lines or equi-
potential surfaces. Consider the equation of an equipotential
In a word, both V2 = 0 and V2 t/J = 0 define an irrotational
line given above in the case of a two-dimensional flow:
motion which satisfies the continuity principle. But
o o V2 t/J :f:. 0 is not compatible with the existence of.
-dx + -dy = 0
ox oy
11-3.3.2 The simplest example of motion in which the
or velocity potential may be used is the two-dimensional
u dx + v dy = 0 uniform flow for which the velocity potential function is

It is deduced that the slope of an equipotential line is ljJ =Ax+ By

dyjdx = - (ujv), which is normal to the slope of a streamline as shown in Fig. 11-3.
(see Section 11-2.2.1). More generally, this may also be The velocity components at any point are
deduced from the fact that oljyjox, o/oy, o/oz are the
direction cosines of the perpendicular to the surface defined o
U=-=A
by= K. ox
 Chapter 11 : Flow Pattern, Stream Function, Potential Function

This momentum equation is often introduced as a free- 121

surface condition for which p is constant. But, in that case,
another unknown must be introduced: z = YJ(x,y,t), which
is the equation of the free surface.
The boundary conditions at a fixed boundary are
o/on = 0. They indicate that the velocity component
in a direction perpendicular to the boundary is zero.
An irrotational flow under pressure is determined, at
least in relative value, from continuity V2 = 0 and fixed
boundary condition 8/on = 0 only.
Figure 11-3 Uniform flow defined by a velocity
potential function.

11-4 Steady Irrotational Two-

That is the very same flow as that given by the stream Dimensional Motion, Circulation of
function
Velocity
1/1 = Ay- Bx
The equipotential lines are given by equating to a constant 11-4.1 A Review, An Example, Polar
value K Coordinates
=Ax+ By= K 11-4.1.1 Table 11-1 establishes a parallel between the
They are straight lines of slope stream and potential functions. The conditions summarized
in this table are those satisfied by irrotational two-dimen-
y A sional motion.
X B These characteristics involve others that have permitted
the development of a number of very versatile tools to
It may be noticed that these equipotential lines are per-
study steady, irrotational, two-dimensional motions. For
perpendicular to the streamlines (see Section 11-3.2).
this reason, this kind of motion has taken on great impor-
11-3.4 General Remarks on the Use of tance in hydrodynamics and also in engineering practice
the Velocity Potential Function since many three-dimensional motions can be analyzed
successfully by neglecting the vertical or one horizontal
Introduction of rather than u, v, w in the basic component. For example, the flow in a wide river when the
momentum equation and continuity relationship reduces
backwater curve effect is small, or the flow toward a well,
the number of unknowns from three (or two in the case of a may often be considered as two-dimensional motion.
two-dimensional motion) to one. However, the order of (The reader is referred to Chapter 2 to distinguish when
differentiation is increased by one degree. Then the system a flow may be considered as irrotational and when the
of equations to be solved has the general form method described below can be used.)
Continuity: V2 = 0
Momentum: 11-4.1.2 The simple example of uniform flow has already
been shown. Another example of irrotational two-dimen-
2 sional flow is that defined by the stream function: ljJ = xy.
a
pat + 2P [(a)
ax + (a)2
ay + (a)2]
oz + p + pgz = f(t) Giving 1/1 various constant values, it can be seen that the
 Part 2: Some Mathematical Treatments of the Basic Equations

122 Table 11-1 Stream and potential functions Therefore, a velocity potential function exists.

Stream function Potential function o oljl

u-----x
o oljl
-ox- oy- v=-=--=-y
oy ox
Continuity Irrotationality
au+ av = 0 au- av = 0 Hence
ax ay ay ax
permit definition of <P = Jx dx tx
= 2 + f(y)
The stream function t/J The velocity potential
function cp
The streamlines are defined by The equipotential lines are <P = JY dy = -ti + f(x)
defined by
dt/J = 0 dcp = 0 It is easy to verify that
t/J=K c/J=K
The velocity components are
at/1 acp satisfies these two conditions.
U=-
ay U=-
ax The equipotential lines defined by = constant form a
atjJ acp family of rectangular hyperbolas which are always per-
v=- pendicular to the streamlines.
V= - -
ax ay
v = ( ~~) 11-4.1.3 Before studying some typical flow patterns, it is
useful to establish some fundamental formulas in polar
dn is the part of an equipotential dS is the part of a streamline coordinates. Referring to Fig. 11-5, it is seen that
line defined by defined by
dcp = 0 dt/J = 0 V = u + v = v, + Ve
dn is normal to the streamlines dS is normal to the and
equipotential lines
v, = u cos e + v sin e
Irrotationali ty is expressed by Continuity is expressed by Ve = - U sin 8 + V COS 8

V2 t/J = 0 V2 c/J = 0
Also
X= r COS 8
streamlines are represented by a family of rectangular y = r sine
hyperbolas that represent a flow toward a plate per-
pendicular to the incident motion, as shown in Fig. 11-4. ox ! ox= -sine
Such a motion is irrotational because u = x, v = - y, and or= cos e roe
1 oy
~~=sine --=cos
roe
e
 Chapter 11 : Flow Pattern, Stream Function, Potential Function

123

Figure 11-4
Flow toward a plate

Now consider t/J in terms of these coordinates. Introduce q; become

In the same manner, the equations for
t/J into the above equations to obtain aq; ax aq; ay aq;
v =--+--=-
' ox or oy or or
1
V, = -;: [a"'oy (r COS 8) + ot/J . 8)
OX (- r Sill J and

_ 1 [ot/J
-- -- - ox]
oy+ot/J _-
- - 1 ot/J
- Vo = -;:
1 [aq;OX (- r . 8) + aq;oy (r
Sill COS 8)
J
r oy o8 OX o8 r o8
and = ~ [aq; ax + aq; oy] = ~ aq;
r ox o8 oy o8 r o8.
V0 = - -
. 8 - -ot/J COS 8
ot/J Sill
oy ax Finally
ot/J 1 aq;
ot/J ay ot/J ax ot/J Vo =--=--
or r o8
oy or OX or or
By a similar calculation, the condition for irrotationality
of a two-dimensional flow
Figure 11-5 Polar coordinates notation.
2( = ou- ov = 0
y oy ax
becomes in polar coordinates
2( = !o(rvo) - ! ov, = 0
r or r o8
Note when v, = 0, rv 0 = constant. It is a free vortex.
 Part 2: Some Mathematical Treatments of the Basic Equations
124
11-4.2 Elementary Flow Patterns
11-4.2.1 Many cases encountered in engineering practice
closely approximate some standard flow patterns. A great
number of them are obtained by a combination or a trans-
formation of three elementary flow patterns. These three
basic patterns are, as described in Fig. 11-6.
1. Uniform flow, studied as an example in Sections
11-2.2.2 and 11-3.2
2. Radial flow: source or sink
3. Circular flow or vortex flow, which is an irrotational
flow with a circulation of velocity. If one or more
vortices are included, the resulting complex flow
pattern is still irrotational. However, the circulation
of velocity may not be zero if the area defined by the
path of integration includes one vortex.
11-4.2.2 Some examples of elementary combinations
of flow patterns without circulation are, as shown in
Fig. 11-7,
Figure 11-6 Basic flow patterns.
SINK VORTICES
SOURCE
1. One source and one sink
2. A doublet: a source and a sink at the same point
3. Flow past a half-body: a source and uniform flow
4. Flow past a cylinder: a doublet and a uniform flow
5. Flow past a long body: (Rankine body) or a stream-
lined fixed body: a source, a sink, and a uniform
flow, or a source, a series of sinks, and a uniform flow
11-4.2.3 Some elementary combinations of flow patterns
with circulation of velocity are as shown in Fig. 11-8
1. Spiral vortex: sink and vortex
2. Flow past a cylinder with circulation of velocity
3. The flow past a cylinder with circulation may be
transformed by a conformal mapping operation to the
flow around a wing: this is the theory of an airfoil (see
Section 11-6.4).
11-4.2.4 A source is a flow that moves radially outward
from a point assumed to be infinitely small (Fig. 11-6).
A sink is a flow radially inward to a point.
In practice, such a flow is fairly well represented by the
flow through a porous medium toward a well of small
diameter, insofar as the vertical component is small, i.e.,
insofar as the curvature of the water table is small.
However, as previously mentioned, its main interest
lies in the fact that complex flow patterns usually en-
countered in engineering practice may be obtained by a
combination of sources, sinks, and other elementary kinds
of flow.
Let Q be the discharge of the source. The components
of velocity at any point are v6 = 0 (for the purpose of
symmetry) and
Q a 1 ol/f
v =~=-=--
r 2nr or r ()()
Stream function and velocity potential function are
given by direct integration. They equal, respectively,
1/1 = _g_ ()
2n
<P=~lnr
 Chapter 11 : Flow Pattern, Stream Function, Potential Function

125

FLOW PAST A HALF BODY

(c)
Figure 11-7
(a) (b)
Examples of
combination of basic flow
patterns without
circulation.

FLOW PAST A BODY

(e)
FLOW PAST A CYLINDER

(d)

Equipotential lines, given by = constant, are circles the velocity distribution is governed by the law v9 r =
(r = constant). Streamlines, given by 1/J = constant, are constant = K/2n. It may be seen that when r tends to zero,
straight radial lines (8 = constant). Changing Q to - Q v9 tends to infinity. Such a motion is irrotational.
gives the velocity potential function and stream function As there is no radial flow
of a sink.
It is easy to verify that the velocity potential function _ o _ ~ oi/J _ 0
v,- or -roe-
of a three-dimensional source, where V = Q/4nr 2 is
= - (Q/4nr). In this case, the equipotential surfaces Hence, one obtains
= constant are spheres (r = constant).
v9 = V = -
K 1 o
= -- = - -
ol/f
11-4.2.5 A vortex is a flow in which the streamlines are 2nr r oe or
concentric circles (Fig. 11-9). In a "forced vortex" water
which yields
turns as a monolithic mass, the velocity being proportional
to the distance from the center (Section 2-4.1 ).
The flow under study is a "free vortex." In a free vortex 1/1 =- J!S_
2nr
dr = - K In r
2n
 Part 2: Some Mathematical Treatments of the Basic Equations

126

SPIRAL VORTEX FLOW PAST A CYLINDER

Figure 11-8
la) (b)
Examples of.fiow combinations with
circulation of velocity

(c)

The flow pattern is very much the same as that of a source 11-4.3 Circulation of Velocity
or a sink, but the streamlines and equipotential lines are -Definition
interchanged. As the flow is irrotational, the Bernoulli Circulation is a mathematical concept on which the
equation may be applied throughout the fluid: theories of wings, airfoils, blades of pumps or turbines,
v2 p* propellers, fans, the Magnus effect (which causes deviation
-. + - = constant of a tennis ball), some sand motions in a flow, etc. are
2g 05
This yields based.
K2 p* Circulation is given by the integral along a closed curve
-8 2 2 + -= = constant S of the tangential velocity component V s
n r g w
It is interesting to note that when r tends to zero, p* /05
tends to - oo. Hence, the presence of vortices in a flow
is a very important cause of cavitation when air is not
admitted into the core from a free surface. Then capillarity It can be demonstrated that r is equal to zero in an
forces take on importance when r -+ 0. irrotational flow.
 Chapter 11 : Flow Pattern, Stream Function, Potential Function
Figure 11-9 Vortex.
There is an exception if the closed curve is around a point
which is the center of a vortex. Then
r = Ve2nr
and since
K considered as the round nose body of a pier, for example.
Ve=~
2nr
Such a flow is called irrotational with circulation. The
circulation along a closed curve in a rotational flow is
generally different from zero, and it may be demonstrated
that when the closed curve is around an elementary area
dx dy, dr = ( dx dy. It can also be demonstrated that the
circulation r is equal to the flux of the vector rotation of
components (, IJ, ~ through the considered area limited
by the curve S. Only the definition of the circulation is
given here, since it is important to know at least its defi-
nition. Its use requires further study beyond the scope of
this book.
11-4.4 Combination of Flow Patterns
11-4.4.1 As previously seen, a great number of very
complicated flows are obtained by simple addition of the
three basic flow patterns studied in the previous paragraphs:
(1) uniform flow; (2) radial flow: source or sink; (3) circular
flow: vortex.
Examples are first given; then the conditions for adding 127
flow patterns, velocity potential functions, or stream
functions are analyzed by consideration of the boundary
conditions. Also, some more general considerations of the
methods of calculation in hydraulics are given.
11-4.4.2 It has been seen that a uniform flow may be
defined by 1 =Ax, or t/1 1 = Ay. A source may be defined
by 2 = (Q/2n) In r, or t/1 2 = (Q/2n)8. Their addition gives
the pattern defined by the velocity potential function
Q
<P = </1 1 + </1 2 = Ax + -In
2n
r
and the stream function
This flow pattern is presented in Fig. 11-10. It may be noticed
that a central streamline completely separates the source
from the outside part of the plane. This streamline may be
In elevation, the upper half of the flow pattern might be
regarded as the flow of wind above a hill.
Streamlines and equipotential lines may be obtained
graphically from the two basic flow patterns. It is sufficient
to add a value t/1 1 = K 1 (or 1 = K'1 ) to a value t/1 2 = K 2
(or 2 = K2) in such a way that K 1 + K 2 (or K'1 + K2)
are always equal to a constant value K.
For example, the intersection of t/1 1 = 4 with t/1 2 = 4
gives t/1 = 8. The intersection of t/1 1 = 3 with t/1 2 = 5
also gives t/1 = 8. The line joining all the intersections for
which t/J = 8 is the streamline marked t/J = 8. The drawing
is very simple when the same interval J..t/f (or J..) is chosen
in the two elementary flow patterns. In the case of Fig. 11-10,
this interval J..t/f is unity.
11-4.4.3 Similarly, it can be demonstrated that one source
and one sink of same intensity and located at the same point
form a doublet defined by the stream function t/1 1 =
- (K sin 8/r). The addition of a doublet with a uniform
flow t/1 2 = ur sin e gives a streamline in the shape of a
 Part 2: Some Mathematical Treatments of the Basic Equations

128

a_.---~.::- 6

--~~~~--~~--~~~~--~~~--~~~----~~----5
Figure 11-10

----~~~--~~~~~~-~~~--~~~~===4 Graphic method of addition offlow patterns.

cylinder. Hence the outside flow pattern is considered as the is nil. This result is general. The total force exerted by
flow of a perfect fluid around a cylinder. The stream function uniform stream of a perfect fluid on a submerged body,
for flow around a cylinder is without circulation of velocity, is nil. This is the paradox
of d'Alembert.
K sin(}
1/1 = - + Ur sin (}
r 11-4.4.4 If 4Yt. 2 , ... , ., ... are solutions of V2 = 0,
or any combination = 1 + 2 + + . + is also
1/1 = u(r - ~2 ) sin (} a solution of V2 = 0 and is therefore a possible flow
pattern. A similar rule exists for the stream function 1/1, and
where R = (K/U) 112 and U is the velocity at infinity. It can solutions of V2 !/J = 0. However, boundary conditions
be demonstrated that R is the radius of the cylinder. The have to be satisfied, and will not necessarily allow the simple
potential function is found equal to addition of elementary solutions to yield the final potential
(or stream) function.
= - u(r + ~2 ) cos e 11-4.4.5 Consider the flows presented in Fig. 11-11.
The velocity distribution around the cylinder is In the first case of a flow under pressure, an addition of
solutions characterized by velocities vl and v2 at a given
18 . point does not change the flow pattern because the pattern
V = Voir=R =-;:a(} = 2U Sill(}
does not depend upon the absolute value of the velocity.
and the pressure distribution is In the second case of flow with a free surface, the flow
pattern is changed because the slope of the free surface
(p - Po:J = tpU 2 [1 - 4 sin 2 (}] changes with V. The solutions cannot be added, as they
where p oo is the pressure at infinity. It can be verified that depend upon the absolute value of the velocity. This stems
the net pressure force on the cylinder from the fact that the flow depends upon a non-linear
equation: the momentum equation; or more specifically the
f"/4 Bernoulli equation, in which the elevation of the free surface
F = 4 Jo p cos (} R d(}
is related to the square of the velocity. The first flow pattern
 Chapter 11 : Flow Pattern, Stream Function, Potential Function

Hence, the flow pattern and the velocity field on one side, 129
and the pressure distribution and free-surface streamlines
on the other side, cannot be calculated independently by

~~ successive steps as in the previous case. The flow pattern

depends upon the absolute value of the velocity, which
~~ may be known only by a combination of linear equations
~
~
(continuity) with the nonlinear momentum equation. The
assumption of irrotationality may be introduced in the
~I I
f?Z????? ~~l~II!
2 ? ????? momentum equation, but this does not make the free
surface condition linear. The considerations given above
Figure 11-11 The flow pattern depends upon the shape of
the free surface.
lead to some more general remarks on the importance of
the boundary conditions.

under pressure may be drawn directly from the fixed bound- 11-5 Reflections on the Importance of
ary, which defines two streamlines. This flow pattern Boundary Conditions
depends only on linear relationships:
11-5.1 New Theoretical Considerations
1. The continuity: au + av = 0 or vz = 0 on the Kinds of Flow
ax ay
From the previous considerations, it is seen that in any
2. The irrotationality: au- av = 0 or V2 t/J 0 kind of flow the method to be used to determine the flow
ay ax =
pattern depends upon the kind of boundary conditions
3. The boundary and upon the assumption of irrotationality. From this
acp = 0 point of view, two major categories of motion may be
condition: an distinguished that are encountered in all methods in hydro-
This boundary condition involves the continuity only. dynamics: analytical, numerical, and graphic methods,
This flow pattern does not depend upon the absolute or methods based on an analogy. The major categories are
value of velocity, only upon its relative value. In a word, the on one hand the irrotational motions under pressure
solution for the flow pattern under pressure within given and slow motion, and on the other hand, the free-surface
boundary is unique. flow and flow with friction force.
Then the pressure distribution may be calculated in a
final, independent step by application of the momentum 11-5.2 Irrotational Motion under
equation in the form of the Bernoulli equation.
Pressure and Slow Motion
In the second case, the flow has a free surface. This free The first category includes all irrotational motions under
surface is unknown and must be calculated. Both the pressure, or motions considered as such, and slow motion
nonlinear momentum equation and the continuity equation in which the quadratic terms are negligible.
must be taken into account.
The boundary condition at the free surface p = constant 11-5.2.1 In the case of flow under pressure, the stream-
involves a force and must be introduced in the momentum lines at the boundary are fully determined since they are
equation to calculate the shape of the free-surface stream- coincident with this boundary. The boundary conditions
lines. In turn, this shape has an effect on the flow pattern. are expressed to satisfy the continuity principle, that is,
 Part 2: Some Mathematical Treatments of the Basic Equations
130 that the velocity is tangential to the boundary. The flow
pattern depends completely upon linear equations only,
expressing the continuity and the irrotationality principles,
after which the flow pattern is relatively easily known.
The velocity field gives the relative value of the velocity.
The absolute value can be known when the velocity is
determined at one point, either given by a boundary con-
dition or calculated by application of the momentum
equation at this boundary.
Finally, the pressure distribution is determined from the
knowledge of the velocity at any given point by application
of the momentum equation.
11-5.2.2 In the case of slow motion, the motion is mathe-
matically considered as infinitely small, even with a free
surface. Hence, all the quadratic terms may be neglected
and the momentum equation becomes linear. The free
surface is considered to be known at the beginning and is
denoted by a horizontal line. In that case, various solutions
of flow patterns may be added.
For example, if 1 is the velocity potential function of a
periodic gravity wave at the first order of approximation,
that is when the convective inertia term is neglected, and if
2 is the potential function of another wave traveling in the
opposite direction, 1 and 2 are determined by a system
of linear equations, as will be seen in Section 16-3.
Hence, = 1 + 2 is the potential function of the
resultant motion.
11-5.3 Free-Surface Flow and Flow with
Friction Forces
The second category of motion includes all motions
with free surfaces or the motions for which the friction
forces have a nonnegligible effect, causing the motion to be
rotational.
11-5.3.1 The free-surface condition involves a force
(p = constant). This force can only be inserted in the
momentum equation, which is an equality of force.
The flow pattern is now determined from nonlinear
boundary conditions as well as from the linear continuity
relationship.
Difficulties arise from this nonlinearity and the fact
that the free surface is unknown. 1 and 2 as defined in
Section 11-5.2.2 may also be calculated at a higher order of
approximation from the nonlinear free-surface conditions,
which take into account the convective inertia term. How-
ever the solution , representing the two nonlinear waves,
cannot be obtained by simple addition of the new solutions
1 and 2 given at a high order of approximation. It must
be calculated from the basic equations. Similar consider-
ations prevail in the case of irregular waves traveling at
different velocities in the same direction. There is nonlinear
interaction.
In conclusion, in order that the velocity potential
functions and the stream functions may be added, they must
depend upon a linear and homogeneous equation only. Also,
the boundary conditions should be homogeneous.
11-5.3.2 Similarly, a friction force (resulting in a rota-
tional term different from zero) gives a boundary condition
V = 0. Such a boundary condition must also be introduced
in the momentum equation. The nonlinearity of the momen-
tum equation, caused by the convective inertia term, is the
major cause of difficulty in studying this kind of flow.
These mathematical difficulties show the importance of
the irrotational motion under pressure and of the slow
motions in hydrodynamics, even if they only represent very
approximately the natural conditions.
11-6 Flow Net
11-6.1 Flow Net Principle
The flow net is a family of equipotential lines and a family
of streamlines representing a complete two-dimensional
flow pattern, as was shown in Fig. 11-4.
The equalities V = ocpjoS = oljljon for a finite difference,
take the form V = f'!..1 f'!..S = f'!..ljl If'!..n.
First, f'!..ljl is chosen to be a constant in the complete
velocity field, which means that the discharge f'!..Q between
 Chapter 11 : Flow Pattern, Stream Function, Potential Function
I I
Figure 11-12 Flow net started in radial and
uniform flows.
two adjacent streamlines is the same (/11/f = f..Q) (see
Section 11-2.3).
Second, the interval /1 is chosen to be equal to !11/J,
which leads to f..S = f..n. f..S is the streamline element,
whereas f..n is an equipotential line element with right-angle
intersections. Hence, f..S and f..n are the two sides of a
curvilinear square, which tends to be an exact square
when f..S and f..n tend to the infinitessimals dS and dn.
This characteristic of a two-dimensional irrotational
flow permits one to draw a complete flow pattern as a mesh
of squares, as shown in Fig. 11-12. At any point, the velocity
direction is given by the streamline. The magnitude of the
velocity given in relative value, is inversely proportional
to the square sides.
The graphic procedure for construction of a flow net
depends on whether the flow is under pressure or with a
free surface.
11-6.2 Flow under Pressure
11-2.6.2 The first case in which the flow net method may
be of very great use is for flow between two fixed boundaries,
which corresponds to a flow under pressure.
A flow net around a solid body with well-determined
boundary conditions at infinity follows the same rules of
construction. It should also be noted that many flows at the 131
free surface, such as in a wide and relatively shallow river
in which rotation is about a horizontal axis, may be defined
by a two-dimensional velocity potential function and
studied by the flow net method (see Section 2-6.4). How-
ever, wave effects or backwater curves must be neglected.
All these types of flow are determined by the same method.
11-6.2.2 The flow pattern is started in the regions in which
the velocity distribution is known, such as in a uniform or
radial flow (Fig. 11-12). Then a number of streamlines
are selected as a function of the desired accuracy, taking
into account that this number could easily be increased
in a given area if a greater local accuracy is required. Then
the equipotential lines are drawn intersecting the stream-
lines (including the boundaries) perpendicularly, and
forming squares with the streamlines.
The simplest method of checking the correctness of the
drawing is to draw the diagonal lines of the square mesh.
These diagonals should, themselves, form smooth curves
that intersect each other perpendicularly. This is Prasil's
method, as demonstrated in Fig. 11-12 (see also Fig. 11-14).
If these diagonals do not intersect, a second drawing is made
by superimposition of transparent paper to correct the
first mistakes, repeating the process until the desired result
is obtained. As a rule, three successive drawings are sufficient
to obtain an accurate flow net by trial and error.
11-6.2.3 The limits of validity of the flow net method in
studying flow under pressure are the same as those imposed
by the assumption of irrotational motion. That is, the flow
net method may be applied to study short convergent flow,
or flow through porous medium when the Reynolds
number is smaller than 1.
Divergent flow causing separation and wakes, long
structures for which the friction forces cause the motion
to be rotational, and unsteady motion cannot usefully be
studied by the flow net method. In the case of a wake,
a flow net method may be used if the separation line is
determined (see Fig. 11-13). The pressure and velocity
Part 2: Some Mathematical Treatments of the Basic Equations
--J..i 1--b
_J I.-a
WAKE Llcp =a: b =....
Figure 11-13 Wake.
are then considered to be constant along this line. Its
determination is relevant to the method for flow with a
free surface, which is the subject of the next section.
11-6.3 Flow Net with Free Surface
11-6.3.1 When the free-surface boundaries are known by
previous experiment, the same method as that explained
to construct a flow net under pressure may be used. More-
over, a free-surface condition is given that determines the
distance between equipotential lines.
Three cases may be distinguished.
1. Flow through porous medium: the vertical distances
between successive equipotential lines, following
the rule /1 = constant, are constant as has been
shown in Section 9-2.2.3 (Fig. 9-4).
2. Horizontal high velocity flow: a flow through an
orifice or from a gate, with a contraction and under a
high head (Fig. 11-14). In this case, by application
of the Bernoulli equation, V = [2g(H + z)J 1i 2 , H
being the total head, whereas z refers to the exact
elevation of any point under consideration above
the level downstream of the gate. In many cases, z
is always small compared with H, and V is con-
sidered to be a constant at the free surface. Hence,
the distances between equipotential lines are equal.
In a word, such a flow is determined by considering
that the gravity force in the downstream part of the
gate or the orifice is negligible.
3. Flow with vertical velocity component (over a weir):
the velocity at the free surface varies with z. According
to the Bernoulli equation, V = (2gz) 1 12. Hence, the
distance between the free surface and the first stream-
line is given by (z is measured downward herewith)
11 = constant = fln = flS
V (2gz)lf2
11-6.3.2 When the free surface is not known from previous
experiments, the continuity and the momentum equations,
which give conditions such as those presented in (1), (2),
and (3) above, should be sufficient to determine the free
surface streamline and the complete flow pattern. The
solution proceeds as follows: A tentative streamline is
drawn intuitively. Then the distances between equipotential
lines are calculated as shown in Section 11-6.3.1. The flow
net drawn on this basis must be found to be consistent
with the given fixed boundary. If it is not consistent, a
second trial is made by modification of the free surface,
and so the solution proceeds. It is easy to conceive that
such a trial and error method is tedious and inaccurate.
Therefore, although such a procedure is theoretically
possible, it is unrealistic to attempt to determine a flow
Figure 11-14 Horizontal high-velocity flow.
H
constant ..........
,-;;:---- = consfonf
v 2gz
 Chapter 11 : Flow Pattern, Stream Function, Potential Function
net with a free surface without an experiment. This is even
more true when it is a flow between two free surfaces such
as a free-falling jet. Most often the necessary experiments
for determining the free surface are self-sufficient for
practical engineering purposes. Because a model must be
built, it can also be used for measuring the pressure distri-
bution; and the determination of the flow net is then a purely
academic exercise.
It is more accurate to determine the flow pattern by a
numerical scheme and computer. However, the pro-
grammed numerical scheme is based on the same trial-
and-error method.
11-6.4 Other Methods, Conformal
Mapping
11-6.4.1 There are a number of methods for drawing a
flow net. All of them are based on the same principles, and
a similar difficulty is encountered in the determination of a
free surface. The most satisfactory solution is found by
using a trial-and-error numerical method with a computer.
The relaxation method is also based on numerical calculus.
An analogy with an electric field is very often used. In this
case, the analogous equipotential lines between boundaries
at different voltages are measured directly. This method
is used with liquid resistance mesh, wetted earth, etc.
The relaxation method and electrical method may be
easily extended to three-dimensional irrotational flow.
Another analogical method is based on the fact that the
mean motion of laminar flow at constant thickness may be
considered to be irrotational (see Section 2-6.4).
Finally, because the mean motion of laminar flow
through a porous medium is irrotational, it is very easy
to use the analogical method to study any two-dimensional
or three-dimensional patterns.
11-6.4.2 It is beyond the scope of this book to develop
the powerful mathematical tool of conformal mapping
for studying two-dimensional irrotational flows with or
L
y
133
I
I
I
X
Figure 11-15 Notation for conformal mapping.
without circulation of velocity. Only the principle is
described below.
Conformal mapping (Fig. 11-15) is based on the use of
complex numbers and functions of a complex variable.
Two complex planes are used. In one, W = cp + it/1,
whereas in the other Z = x + iy = reiO. A relationship
between the two planes is developed such that W = f(Z).
A conformal mapping operation consists of establishing
a relationship between each point of a given flow pattern
in the x, y plane and a point of another flow pattern in the
cp, t/1 plane. The first is often the real flow under study;
the second is often a uniform or a simpler flow pattern.
Successive conformal mapping operations may also be
done in order to pass step by step from a very complex
flow pattern to a uniform flow.
Conformal mapping can also be used for determining
free streamlines (Schwartz-Christoffel transform). However,
its application requires that gravity forces be neglected.
11-6.4.3 For example, consider the transformation
where W = cjJ + it/1 is the equation for a uniform flow in
the W plane, i.e., in the system of cp, t/1 axes. This flow is
parallel to the cjJ axis,. and the streamlines defined by
t/J = constant are perpendicular to the t/1 axis.
The above relationship is the transformation of a flow
around a cylinder of radius R into a uniform flow. This
can be seen when the real and imaginary parts are separated.
 Part 2: Some Mathematical Treatments of the Basic Equations

134 This is easily done by substituting Z = rei 0 into the trans- Flow past a cylinder with 2
formation and using the relationship eiO = cos (} + i sin e. circulation of velocity:
W = U( Z + -R ) - !__
K In Z
Z 2n

w =+it/! Flow at a wall angle: W= Z",

= u(reiO + ~2 e-iO)
PROBLEMS

= u(r cos(} + ~2 cos(} + ir sin(} - i ~2 sin(})
= u(r + ~2 ) cos (} + iU(r - ~2 ) sin (}
Then it is seen that the potential function is
11.1 Draw a square mesh in a two-dimensional bend such
as shown in Fig. 11-16 and calculate the relative pressure
distribution on both boundaries and along the streamline
starting from point A at the center of the upstream pipe.
11.2 Give the expression for the Navier-Stokes equations
as a function of the stream function t/J(x,y) in the case of
two-dimensional motion

11.3 Demonstrate that the velocity potential function for

and the stream function t/1 is a three-dimensional source is =- (Q/4nr).

11.4 Determine the stream function and the potential

function for a uniform flow of velocity V inclined at an
angle r:x with the X axis.
These two functions are those of a flow around a cylinder
(see Section 11-4.4.3, vide supra).
Figure 11-16

11-6.4.4 The following transformations can be studied 1------- 6a

by using a similar approach:
a/2
Uniform flow: W =(a+ ib)Z A a
a/2
Source at Z =A: W= ~ln(Z- A)
iK
Vortex as Z =A: W = - -In (Z - A)
2n
1
Spiral vortex at Z = A: W = 2n (Q- iK)ln (Z- A)

Source at -A, sink at +A:

Q Z+A
W =-In---
2n Z- A 3a

Flow through an aperture: Z = cosh W

 Chapter 11 : Flow Pattern, Stream Function, Potential Function

11.5 Sketch the streamlines and equipotential lines for a 135

flow past a cylinder of radius R. Determine the corre-
sponding stream function. v ;v
+
R
dR

11.6 Study the various characteristics of a flow defined

by the stream function l/J = - x 2 Determine if such a
flow is rotational, and calculate the vorticity. Is the fluid Figure 11-17
compressible? Plot the streamlines and the equipotential
lines.
and demonstrate that streamlines and equipotential lines are
11.7 Consider a uniform flow in the positive X direction. circles.
The velocity varies linearly from V = 0 at y = 0 to V = 10
ftjsec (3.048 m/sec) y = 10 ft (3.048). Determine the expres- 11.13 Demonstrate that
sion for l/1.

11.8 Draw the flow pattern from a source to a sink by

graphical means.
where
11.9 Consider a flow around a cylinder defined by the vz p
potential function <jJ = - U(r - (R 2 /r)) cos e. At which H=-+-+z,
2g pg
distance is the fluid velocity disturbed by the cylinder by
more than 50%, 10%, and 1%? Sketch these three lines V is the particle velocity, r the circulation of velocity,
of influence around the circle. dR an element perpendicular to the streamlines, and dl
an element of streamlines, as shown in Fig. 11-17.
11.10 Consider a free surface sink vortex in which the
vertical component of velocity will be neglected. Calculate 11.14 The stream function for a flow past a cylinder with
the elevation of the free surface 17(r). circulation of velocity is

11.11 Consider the potential function l/1 =

2
U( r - -R ) sin r In r
e - -2n
r
Q K
<jJ =-In r + -e where r is the circulation. Determine the position of the
2n 2n
stagnation points on the cylinder as a function of r.
Calculate the stream function and the general equations Demonstrate that the total force exerted by the flow per
for equipotential lines and streamlines. Draw the corre- unit length of the cylinder is F = pUr.
sponding flow pattern assuming that Q = K and Q = !K
successively by means of graphic superposition. 11.15 Calculate the potential function for a flow past a
"Rankine" body. The stream function is, as shown in
11.12 Demonstrate that the potential function of a doublet is Fig. 11-18
<P = K cos e
r
Part 2: Some Mathematical Treatments of the Basic Equations

Figure 11-18

Determine the shape of the Rankine body and calculate

the pressure around it in terms of the value of the pressure
at infinity Poo

11.16 Demonstrate by finite differences that, in an irrota-

tional flow, l/1 1 [the value of the stream function at a point
(1)] given by
l/11 = i(l/Jz + l/13 + l/14 + l/Js)
The subscripts 1-5 refer to points 1-5, as shown in Fig.
11-19.

11.17 In the case of a flow past an aperture of length 2C

and defined by the conformal mapping transformation
Z = C cosh W where Z = x + iy and W = + il/1, dem-
onstrate that the streamlines in the z planes are defined
by a family of hyperbolas and that the equipotential lines
are defined by a family of ellipses of same foci.

Figure 11-19

3
Chapter 12 12-1 External Forces and Internal
Forces
137

12-1.1 Considerations on Forces

The Momentum The momentum equation F = m(dVjdt) has been expres-
sed in differential form for an elementary fluid particle of
unit volume and mass p (see Chapter 6). It may be recalled
Theorem and that this momentum equation takes the form of the Navier-
Stokes equation, which equates the inertia force of a unit
Its Applications volume with the corresponding applied forces.
The applied forces have been divided into external and
internal forces (see Chapter 5). The internal forces are
caused by pressure and friction. They are, by definition,
vectorially equal to zero and do not contribute to a net
torque on the considered particle. This definition is based
upon Newton's third law stating that action equals reaction.
The external forces are divided into surface forces caused
by pressure and friction, and a body force caused by gravity.
These forces have a total sum different from zero, hence
impart a motion to the elementary fluid particle.
Consider two adjacent fluid particles, as shown in Fig.
12-1. The external forces acting on the two adjacent sides
sum vectorially to zero according to Newton's third law
as previously stated. Hence, only the external forces acting
on the outer limits of this group of two elementary fluid

Figure 12-1 External forces at the boundaries

of two adjacent fluid particles may
be considered as internal forces in
order to study their overall motion.
 Part 2: Some Mathematical Treatments of the Basic Equations

138 particles affect their overall motion. Therefore, considera-

tion of these external forces permits a theoretical analysis
of the combined overall motion of these two particles, but
does not permit an analysis of the relative motion of the
one particle with respect to the other.
Generalizing for a definite mass of fluid composed of
a large number of elementary fluid particles shows that all
internal forces sum to zero and will produce no net torque
on the definite mass of fluid. The overall motion of this mass Figure 12-2 External forces tend to move the pipe
of fluid depends only upon the external forces applied to it. downstream, but they do not work.
Consequently, although this simplification does not permit The head loss is caused by the work of
internal forces, which have a total
a study of the internal motion within the mass of fluid nor sum equal to zero.
of the fine structure of the flow pattern, a study of its overall
effect may be possible.
12-1.3 Field of Application
12-1.2 Considerations on Energy From the previous considerations, it may be deduced
Rather than express the momentum equation F = that, in practice, the difference in application of the mo-
m(dVjdt) for an elementary fluid particle as an equality of mentum equality and the energy equality lies in the emphasis
applied forces and inertia, consider an equality of work on the importance of internal forces, insofar as the pheno-
and kinetic energy: mena being studied are concerned, which is illustrated by a

dV dS = m -dV V dt = d -
F dS = m-
(mV 2
)
number of examples in this chapter.
A considerable number of hydraulic problems are
dt dt 2 simplified by the fact that the sum of the internal forces is
The separation between internal and external forces is zero. It is for this reason that the momentum theorem is so
always possible. Hence often used.
The momentum theorem is used to calculate the overall
I:(Fe dS) + I:(F; dS) = d(m;
2
) effects of a mass of fluid, however complex the flow, without
dealing with the fine structure of the flow pattern. However,
However, in spite of the fact that the total sum of the in- to apply the method which consists of considering only the
ternal forces is zero by definition (I:F; = 0), the work of external forces to calculate the change in momentum
these internal forces does not equal zero, i.e., I:(F; dS) =I= 0. requires a perfect knowledge of boundary conditions at the
To illustrate this point, consider a uniform flow in a extremities of the mass of fluid under study. This point is
pipe (see Fig. 12-2). The external forces acting at the wall- illustrated in Section 12-4.
boundary have a total sum that is different from zero,
thereby tending to move the pipe in the direction of the flow. 12-1.4 Momentum Theorem and
But the existing internal forces sum to zero. However, these Navier-Stokes Equation
internal forces do work and this work is the cause of the The momentum equation for a finite mass of fluid can
head loss. The head loss expresses the transformation of be established in several ways as the Bernoulli equation.
energy lost by friction into heat. One could make use of the basic Navier-Stokes equation
Thus, insofar as the energy equality is concerned, internal by integrating all the forces causing motion of an elementary
forces may not be neglected. particle of fluid mass p to the forces involved in the motion

of a definite mass of fluid m. This is evident, as the Navier-
Stokes equation is the momentum equation for a mass of
fluid of unit volume.
Instead, a direct vectorial demonstration is given for an
arbitrary mass of fluid of finite dimensions. Although the
momentum theorem is used primarily to solve problems in
steady flow, the more general case of unsteady flow is
considered here. This method illustrates the difficulties
encountered in the application of the momentum theorem
to unsteady motion.
12-2 Mathematical Demonstration
12-2.1 Concept
By definition, the product of mass and velocity is
momentum. Hence the momentum of an elementary particle
fluid of mass pis pV. Therefore, the total momentum of a
definite mass of volume D, in which the velocity vector
varies both with time and direction, is JJJvPV dD, where
dD is an element of the volume D.
Assuming a mass of fluid is bounded at a given time t by
a surface A, the same mass of fluid at time (t + dt) will be
bounded by a surface A', quite similar to A. These two
surfaces define three domains: D 1 , D 2 , and D 3 Whereas
D 2 has a finite dimension, D 1 and D 3 are, by definition,
infinitely small, because the interval of time dt is infinitely
small (see Fig. 12-3). Successive values of the total momen-
tum of the fluid in these three domains is calculated.
Figure 12-3 Notation for momentum of a moving
mass offluid.
A
Chapter 12: The Momentum Theorem and Its Applications
139
12-2.2 Change of Momentum with
Respect to Time
The momentum of fluid enclosed in the common part
D2 at time t is JJJv 2 PV dD. At time t + dt, the velocity
becomes V + (oVjdt) dt and the momentum becomes
Hence, the difference or variation of momentum during the
interval of time dt is the difference JSJv 2 (opVjot) dt dD.
Note that the integral is the product of a finite number D 2
times an infinitesimal number (opVjot) dt. Dividing by dt,
the variation of momentum per unit time is JSJv 2 (opVjot) dD.
This term has a zero value in the case of a steady flow
because av jot = 0.
12-2.3 Change of Momentum with
Respect to Space
The momentum of fluid enclosed in domain D 1 (see
Fig. 12-4) at timet is JJJv,PV dD which is dimensionally the
product of a finite number IV I and an infinitesimal
number D 1 . Domain D 1 may be considered as containing
elementary cylinders of base dA and sides parallel to the
velocity vector V.
The volume of an elementary cylinder is dD = dA V, dt,
where V, is the projected value of V on a perpendicular to
dA. It is deduced that JJJv, dD = JSA, dA V, dt in which A1
is the part of A, which defines the limit of domain D 1 .
Figure 12-4 Momentum notation.
v
I
I
I
I
I
I
L2:;:::__..,.-r::;;.-r--..- Vn
 Part 2: Some Mathematical Treatments of the Basic Equations

140 Hence, the momentum of fluid enclosed in D 1 becomes

This is given by a surface integral rather than by a volume
integral throughout the volume D 1 .
The momentum of fluid enclosed in domain D 3 at time
(t + dt) is
ff{/( v + ~~ dt) dD.
The first integral JHv, pV dD is the product of a finite
number lVI and an infinitesimal number D 3 , whereas the
second integral JHv,
p(oVjot) dt dD is a product of two
infinitesimal numbers 1(8V/8t) dtl and D3 Hence, this
second integral may be neglected.
A process of calculation similar to that just demonstrated
above shows that
where A 2 is the part of A that defines the limit of domain D 3 .
Now the difference of momentum between domains D 3
and D 1 at time (t + dt) and time t respectively, caused by
the variation of velocity with respect to space is
12-3 Practical Application of the
Momentum Theorem-Case of a
Stream Tube-Specific Force
12-3.1 Application to a Stream Tube
The first step in the application of the momentum
theorem consists of choosing the limits of the mass of fluid
to which the momentum theorem can be applied. These
limits are chosen in a section in which the boundary con-
ditions are well known, i.e., fixed boundaries or cross
sections where the motion may be considered as uni-
dimensional.
Because the momentum theorem is a vector equality, it is
convenient to choose some axes of reference. Generally, the
main flow direction indicates one of the axes to be con-
sidered; the vertical axis may be another one.
Finally, the momentum equality is written by projecting
all the forces involved on these axes of reference.
In many cases of unidimensional flow, only the equalities
of force in the direction of flow are of interest.
For a streamtube flow, as shown in Fig. 12-5, the terms
of the momentum equation, presented in Section 12-2.3,
applied to a fluid mass within cross sections A 1 and A 2

dt f{ pvv,. dA + dt f{,pvv,. dA
take the following forms:
2

because the discharge of momentum entering the domain is

affected by a negative sign, whereas the discharge of
momentum leaving the domain is affected by a positive
sign; considering A = A 1 + A 2 , and dividing by the Figure 12-5 Momentum theorem applied to a
interval of time dt, the difference of momentum per unit stream tube-notation
time or momentum flux is

f{pvv,. dA
Finally, the total change of momentum per unit time with
respect to both time and space is equal to the sum of the
external forces. Therefore

LFe = JJL 0 ~~ dD + JLpVV,. dA


An averaging correction factor (1 + Q(/3) should be applied
to the term pQV. This correction is neglected in the following
(see Section 10-3.1).
The external forces are composed of pressure forces,
gravity forces, and shearing forces. For reasons discussed
below, the latter two forces are often neglected, leaving
only the pressure forces to consider.
The pressure forces on sections A 1 and A 2 may be
considered as consisting of two components
1. The sum of the forces caused by constant pressures
Pz and Pb i.e., pzAz- P1A1.
2. Hydrostatic forces applied to the center of gravity of
the cross sections (Fig. 12-7). In the case of a free
surface two-dimensional flow on a horizontal bottom
of depth h, the external forces per unit width are:
The pressure p(x) along the walls limiting the stream
tube could be given at any point by the Bernoulli equation
or by assumptions based on physical observations. If all
the other terms are known, this becomes the only unknown
in the equation, and the momentum theorem provides a
way of finding the value of its integral in the flow direction.
[Note that the pressure force at the limits of the considered
stream tube is expressed along the axis in the mean direction
of flow only (see Fig. 12-6).]
Figure 12-6 Pressure forces on the limit of a
stream tube.
Chapter 12: The Momentum Theorem and Its Applications
When the momentum theorem is applied along a hori- I4I
zontal axis, the gravity force is zero.
The shearing forces, acting at the limit of the considered
volume of fluid, are often neglected in the case of short,
rapidly varied flows, such as given by a sudden enlargement
or a hydraulic jump (see Section 12-3.3.).
12-3.2 Specific Force
The specific force in a cross-section A is the sum of the
momentum flux and the pressure force per unit of weight of
fluid, i.e.
vz p)
(-+-A
g pg
In order to illustrate previous considerations and to
provide a guide for further applications, some examples are
given with an emphasis on all the necessary assumptions
not usually given in elementary textbooks on hydraulics.
12-3.3 Hydraulic Jump on a Horizontal
Bottom
From observation, it is common knowledge that the flow
pattern in a hydraulic jump is extremely complicated.
However, consideration of the external forces only and
change of momentum at the boundaries permits the study
of this complex phenomenon without having to deal with
the complicated fine structure of the flow.
First, the flow limits are chosen in a plane at which the
flow pattern is well known, i.e., far enough from the front of
the hydraulic jump for the mean flow to be parallel to the
bottom (Fig. 12-7).
Figure 12-7 Hydraulic jump-notation.
X
0
Part 2: Some Mathematical Treatments of the Basic Equations

Second, two reference axes are chosen. One axis will

obviously be chosen in the direction of flow OX.
12-3.4 Hydraulic Jump in a Tunnel
The external forces to be considered in the 0 X direction
are Consider the case of a partially open gate in a tunnel sub-
mitted to a high upstream pressure, as shown in Fig. 12-8.
1. The pressure forces at the boundaries, i.e., on the
An air vent is often necessary to avoid cavitation effects.
vertical planes AB and CD, having a total sum in the
When conditions for' a hydraulic jump are satisfied, the
OX direction different from zero. The pressure
water flow acts as an ejector and a quantity of air is sucked
distribution is hydrostatic.
into the tunnel. Because of the head loss in the air vent, the
2. The shearing stresses caused by friction on the
pressure at the free surface is smaller than atmospheric
boundaries, including the free surface, and on the
pressure. Therefore, the external forces must take into
planes BC and AD in a direction opposite to OX.
account this difference in pressure. The simplest solution is
In such a short structure, however, these shearing
obtained using the absolute value P = p + Pa of the
stresses are negligible.
pressure which gives
Hence, the external forces acting in the OX direction are
per unit width

P 2 is greater than the atmospheric pressure because of the

The difference in momentum with respect to time is
Equating the change of momentum with the external forces
leads to
head loss in the downstream part of the tunnel. The mo-
mentum flux in cross section (1) is pQ 2 /A 1 In cross section
(2) it is pQ(Q + Qa)/A 2 because of air entrainment. Qa is
the discharge of air at pressure P 2 ; pQ is the mass per unit
time; and (Q + Qa)/A is the velocity. The mass of air is
neglected, but the air discharge has an influence on the
velocity of the water.

Figure 12-8 Hydraulic jump in a tunnel.

This finally becomes, after some elementary transformations
AIRVENT

--'-------,- J3..f:..Wf!..E_LI NE
1 ERIC PRESSURE-----
This is the hydraulic jump equation. Choosing a second
axis in the vertical direction will give an equality between
the atmospheric pressure force acting on the free surface,
the gravity force, which is equal to the total weight of water GATE
in volume ABCD, and the external force acting vertically
upward on the bottom of the hydraulic jump. There is no
change of momentum in this direction.
 Chapter 12: The Momentum Theorem and Its Applications

143
12-3.5 Paradox of Bergeron
Consider a tank on wheels as in Fig. 12-9. On one side
the pressure distribution is hydrostatic, whereas on the
other side the pressure head is transformed into velocity
head. The difference of total forces acting on both sides of the
tank could be obtained by calculating, successively, a flow
net, the pressure distribution on the two sides, and the forces.
Figure 12-10 Paradox of Bergeron:
However, the momentum theorem gives the total value of
A - (B + C) = E - D.
the force directly as
F = pQV = pCcA(2gh) 112
The same result is obtained in the case shown by Fig.
where Cc is the coefficient of contraction. 12-11. The jet acts on the wall of the downstream tank.
Because of this force and the fact that other forces (such The tank does not move. Finally, consider the two tanks as
as gravity and atmospheric pressure) have a horizontal shown in Fig. 12-12 in which it is assumed that the holes
component equal to zero, the tank has a tendency to have the same cross section. One of the holes is closed by
move in the opposite direction of the jet. This is the principle
means of a plane held in place by the jet from the left tank.
of jet propulsion. The area of pressure ABCD equals the area A'B'C'B' or
Now suppose that water is present outside the tank and ABE+ FDC = E'B'C'F'. Considering the forces on the
that the tank is heavier than the buoyant force (Fig. 12-10).
plate we obtain F L = F R when
In this case, the tank does not move. The force caused by
the jet is equal to the force caused by the very complicated pACA2gzt) = pgAz 2
motion inside the tank. This may be considered as another
i.e.
application of Newton's third law-action is equal and
opposite to reaction-and the momentum theorem must be 2Cdz 1 = z 2
applied to the total mass of water. This is what is known as
With cd = 0.60 for an orifice, it is possible for Zz to be
the Paradox of Bergeron. 1.2 z 1 by the simple insertion of a plate. Shaping the hole

Figure 12-9 Jet reaction principle. Figure 12-11 Paradox of Bergeron:

A - (B + C) = B - E.
 Part 2: Some Mathematical Treatments of the Basic Equations
144
I
I side) of the bend. The projected cross section is A. The force
L_ -tt-------,~-- - -: -+ -. . ,. ~r-tt-- J
Figure 12-12 Another paradox:
AEB + FCD = E'B'C'F.
in the left tank so that cd--+ 1.0, z2 can be made almost
equal to 2z 1 . If the jet returns through 180, z 2 could be
equal to 4z 1 .
All of this may be physically explained by the considera-
tion of external forces (AEB and DCF) transformed into
momentum.
12-3.6 Jet and Intake
Let us consider the case of a curved pipe surrounded by
water as shown in Fig. 12-13. When the water is expelled
from the pipe, the total force exerted on the pipe is F = pQV.
The pipe tends to move in a direction opposite to that of the
jet. For example, if a pool is filled with a flexible hose, the
hose will continuously wiggle. When the water is now sucked
Figure 12-13 Jet reaction compared to intake.
F=pQV F =O
by the pipe, the total force on the pipe is obtained by calcula-
ting the difference of pressure on both sides (inside and out-
is Fe= PeA outside. Inside the bend, the force projected on
A is Fi =piA+ pQV. This is caused by the change of
momentum flux direction in the bend. However, application
of the Bernoulli equation gives
where K is a head loss coefficient at the entrance approxi-
mately equal to unity, as seen in elementary hydraulics.
Therefore
The total force on the pipe is zero. The pipe does not move,
as can be verified by siphoning a pool with a hose. Therefore,
one can now imagine a boat equipped with vertical pipes
bent horizontally, as shown in Fig. 12-14. The pressure
variation at the bottom end of the pipe that results from
heaving will cause the water to move alternately up and
down. (This phenomenon can be enhanced by resonance.)
Even though the previous theory applied to steady flow
and local inertia effects are neglected, it is seen that when the
water goes up inside the tube, the boat does not move,
whereas when the water is expelled, the boat will tend to
move forward. The boat is moved by wave energy.
Figure 12-14 When the water is expelled, a reaction is exerted.

12-4 Difficulties in the Application of
the Momentum Theorem
12-4.1 Difficulties in the Case of
Unsteady Flow
In the case of steady flow, the integral with respect to
volume D(fffv (apVjat) dD = 0) and the momentum theo-
rem becomes simply
LFe = JipVY.t dA
As for external forces, only the boundary conditions for
A appear in this equation because the momentum flux is
given by a surface integral. Hence its application does not
require a knowledge of the fine structure of flow within the
domain D but only on the surface A.
In the case of unsteady flow, the volume integral
fJJv (apVjat) dD is different from zero and requires a
knowledge of the flow patterns within D as a function of
time and space. Thus, the momentum theorem is difficult to
apply to unsteady flow and as such, is less frequently used
except when V(t) can easily be determined.
In the case of steady motion, the momentum theorem
permits the analysis of complex motions. This leads one to
think about the precautions that have to be taken in the
application of this theorem.
One difficulty in the application of the momentum
theorem is in the choice of the boundary and the boundary
conditions. It is impossible to calculate the external forces
without having a rough idea of the flow pattern. This often
requires experimentation or knowledge of similar previous
experiments. Some examples to illustrate these considera-
tions are given below.
12-4.2 Sudden Enlargement
The external forces involved are the pressure forces on
parts (1) and (2) in Fig. 12-15. It is generally assumed that
the pressure p* exerted by wall S is the same as the pressure
pf at the end of the smaller pipe. If the flow is laminar, the
Chapter 12: The Momentum Theorem and Its Applications
145
(I) (2)
Figure 12-15 Sudden enlargement.
streamlines have such a curvature that this assumption is
incorrect (see Fig. 12-16). However, it is known by experi-
ment that when the Reynolds number is greater than a
critical value, the flow enters the wider pipe in the form of a
jet. This jet, often unstable, generates by friction some
secondary currents in the corners.
If the velocity is small enough for the convective inertia
to be negligible, it is true that the pressure distribution at the
cross section (1) is hydrostatic and this usual assumption is
valid. In fact, the eddies caused by friction induce a centri-
fugal force that causes the pressure force to be slightly
greater than the force calculated with the previous assump-
tion. However, the assumption is quite valid for practical
purposes.
12-4.3 Hydraulic Jump Caused by a
Sudden Deepening
This example illustrates a case where it is impossible to
calculate external force without experiments.
Suppose that a channel has a sudden deepening in order
to fix the position of a hydraulic jump (Fig. 12-17). The
external force exerted by the vertical wall changes as shown
in the diagram, depending upon the exact location of the
Figure 12-16 Sudden enlargement. Laminar flow.
1~---
~--
 Part 2: Some Mathematical Treatments of the Basic Equations

q6
Figure 12-17 Hydraulic jump on a sudden
deepening.
jump. Only systematic experimentation gives the factors
that influence this phenomenon, and the corresponding
theory is dictated by the results of this experiment.
12-4.4 Hydraulic Jump on a Slope
The study of a hydraulic jump on a slope involves a body
force-the force of gravity-which definitely influences the
flow conditions (Fig. 12-18). The gravity force to be consider-
ed is the component of the total gravity force in the direction
of the main flow due to the weight of water included between
the two usual limits where the flow is parallel to the bottom.
Hence this force pgAl sin rx is a function of the length of the
hydraulic jump l. This length could be roughly estimated
by experimentation, but it is evident that it cannot be
determined with great accuracy. However, the length of the
hydraulic jump must satisfy the equation
12-4.5 Intake
Consider a free-surface flow as shown in Fig. 12-19.
The rise of the water level in part ABCD may theoretically
be calculated by applying the momentum equation in the
OX direction to the mass of water enclosed in EFBC.
Thus, if one assumes that the flow is perpendicular to the
limit GD:
pQV = pg [ (d + !!d) 2 - d2 ]
~
2 2
which gives !!d = V 2 jg and not V 2 j2g, as may be expected
following a superficial analysis.
However, the practical result is often closer to V 2 j2g
than V 2 jg. This is not because the momentum theory is
wrong, but because the boundary conditions are wrong.
The velocity at section GD is not perpendicular to the cross
section and the momentum theorem should be applied to
the mass BCDJHGEF in order to include the difference
in external forces applied at DJ and HG. Unfortunately,
these external forces cannot be estimated by theory.
12-4.6 Unsteady Flow-Translatory
Wave
As an example of unsteady flow where the momentum
theorem may be used, the case of a translatory wave is
analyzed (Fig. 12-20). The wave is traveling at a velocityW.

pQ(V2 - V1) = pg .
(2hi - 2h~) + pgAl sm rx Figure 12-19 External forces due to the intake
cannot be determined.

~F,_------~A~----;9
I I
I I
Figure 12-18 Hydraulic jump on a slope --r-~. - - " j --+--... X
(I = length ofjump).
- ,__\_: c

>I >II 111171 II 71 :~~~ ~~


A Finally, the momentum balance is
D yields
Figure 12-20 Translatory wave.
It is assumed that the velocities V 1 in front of the wave and
V2 behind it are constant along a vertical plane. Referring to
Section 12-2, applying the momentum theorem to the
fluid within the ABCD control sections, and letting h be
the height represented by line segment BC and h + Ah the
height represented by AD, one finds successively: External
forces:
"LFe = pg[1{h + Ahf - !h 2 ]
Change of momentum with respect to space or variation of
momentum flux:
p f{vv, dA = pV~(h + Ah)- pVfh
Change ofmomentum with respect to time: The momentum of
volume AaED and BcFC remains unchanged. The change of
momentum JSJD [o(pV)jot] dD contains two terms: one
due to the variation of mass, the other due to the variation
of velocity. The variation of mass is due to the addition of
fluid abed, which is pAh W dt, or p WAh per unit of time.
This mass takes a velocity V2 Therefore, the corresponding
variation of momentum is:
dM
V - = pAhWV2
dt
The quantity of fluid within dcFE is Wh dt and this fluid is
subjected to a change of velocity from V1 to V2 Therefore,
the variation of momentum of dcFE is:
Chapter 12: The Momentum Theorem and Its Applications
147
!pg[(h + Ah) 2 - h2 ] = p V~(h + Ah) - p Vfh +
+ pAhWV2 + phW(V2 - Vd
which, when combined with the continuity equation
V2 (h + Ah)- V1h = WAh
h 112
W=V1+(g)
[12 h +h Ah (h +h Ah + 1)] 112
12-5 Momentum Versus Energy
In this section, the field of application of the momentum
theorem is analyzed and compared with the field of appli-
cation of the principle of energy conservation.
As has been seen in Section 12-1.1, the main difference
in application of these two methods lies in knowing the
connection of the internal forces with the phenomena to be
studied.
The momentum theorem is used to study an overall
effect, whatever the complexity of the flow. The principle of
conservation of energy is used to study phenomena linked
to the internal motion and to the very fine structure of the
flow pattern.
It is easy to conceive that the second method will be more
quickly limited in its scope when hydraulic problems are
analyzed. It has b"een seen that a number of assumptions,
such as irrotationality, are necessary before this second
method can be used. These limitations are linked with the
difficulties of integration of the basic Navier-Stokes
equations.
A number of examples will now be used to illustrate the
above points.
12-5.1 Irrotational Flow without Friction
The total thrust of a jet on a fixed or a movable plane, the
force on the bucket of a spillway, or the total horizontal
force exerted on a partially open gate in a tunnel, etc.,
may be calculated by the momentum theorem (Fig. 12-21).
 Part 2: Some Mathematical Treatments of the Basic Equations
q8
F= pQV ~o~ a
C A
In contrast, the precise pressure distribution caused by the
above conditions of flow may be calculated by the following
method:
The flow is assumed to be irrotational and flow net may
be drawn, which gives the velocity distribution. The pressure
distribution is given by the application of the Bernoulli
equation expressing the conservation of energy. It is evident
that the total thrust may also be deduced by this process of
calculation, by an integration of pressure forces Jp dA,
and this could be compared to the result given by the
momentum theorem.
However, the result given by the momentum theorem,
which is obtained without any assumptions, is more exact
provided that the boundary conditions are well known.
12-5.2 Unidimensional Rotational Flow
The momentum theorem may be used to analyze a
number of phenomena, such as the head loss at a sudden
enlargement or a hydraulic jump, for example, whatever
the complexity of the flow.
Combining the force-momentum equation with the
energy-work equation gives the value of the head loss by
calculating the difference in total heads: (V 2 /2g) + (pjw)
+ z. Both the force-momentum equation and the energy-
B
Figure 12-21
Total thrust is given by the momentum
theorem. The pressure distribution is given by
the Bernoulli equation and by a flow net.
work equation are valid to study a diverging flow where
the head loss may be neglected, but there cannot be any
application of the energy-work equation to a sudden
enlargement without the introduction of another term
expressing the head loss, despite the identity of equations
in their differential form.
12-5.3 Mechanics of Manifold Flow
The mechanics of manifold flow is of particular interest
as an illustration of the previous considerations on the
fields of application of the Bernoulli and momentum
equations. Consider a flow as shown by Fig. 12-22. If the
motion is two-dimensional, a first method of analysis
assumes the motion to be irrotational. In such a manner it
is possible to calculate the flow pattern by conformal
mapping, although this method is far from being valid
because of the friction forces.
It is also possible to apply the energy equation to the
two stream tubes which separate into the main conduit
flow and the flow in the lateral, if a term is included to
express energy losses for both stream tubes. Unfortunately,
it is impossible by theory to establish the value of this term.
One can assume that head loss for the lateral pipe stream
tube is that of a bend, whereas for the main flow it is that of
 Chapter 12: The Momentum Theorem and Its Applications

149

PROBLEMS

12.1 Derive the momentum equation by integrating the

I
Eulerian equation to a finite volume. Determine the
/EXTERNAL
1 FORCES correcting terms due to turbulence by integrating the
Reynolds equation to a finite volume.

12.2 Consider the flow through a pipe of radius R 0 ended

Figure 12-22 Manifoldjiow.
by two circular disks of radius R and separated by a small
distance h, as shown in Fig. 12-23. Calculate the total force
exerted by the flow on the lower disk by assuming that the
flow between the two disks is radial and that the total
a sudden enlargement (V - ~) 2 j2g. But systematic experi-
discharge is Q. Take R 0 = h first, and then discuss the case
mental results do not verify this assumption.
when R 0 #h.
A similar simplified approach consists of writing the
momentum equation for the flow at the junction, provided
12.3 Consider a two-dimensional flow such as shown on
that a term is included for the momentum of the flow at an
Fig. 12-24. Draw the corresponding flow net and determine
angle in the lateral or the corresponding unbalanced
the pressure distribution from A to B (assuming no separa-
external force component acting on the walls of the lateral.
tion at C and D), and calculate the total force on AB by
Lack of knowledge of one significant term, makes
integrating the pressure distribution as a function of the
direct application of these methods impossible without
fluid discharge Q. Calculate the same force by application
recourse to experiment.
of the momentum theorem. Explain the discrepancy
between these two results.
12-5.4 Conclusion
Figure 12-23
For any problem dealing with steady flow motion for
which only the overall effect is of interest, the momentum Q

theorem can be applied because of its great simplicity. This

is because the sum of the internal forces is considered to be
zero. However, when using the momentum theorem, one
must be very careful in estimating the external forces and
-1 j ~
boundary conditions. Often an experiment may be necessary
to establish these unknowns.
When more details about the flow characteristics are
required, the system of differential equations giving the h

fine structure of the flow directly must be solved completely.

But the validity of this solution is quickly limited because ~ ~~ ~ I
)::?~><'
of the number of assumptions which must be introduced
in order to simplify the system of equations to be solved. 2 R
Part 2: Some Mathematical Treatments of the Basic Equations

DAM

Figure 12-24
lOOft R = 3.05m
R =lOft GATE
305m

12.4 Demonstrate the following relationship for a hydraulic

jump in a rectangular horizontal channel

B
where y 1 and Yz are the upstream and downstream water
depths, respectively, F 1 = Vd(gy 1 ) 112 is the Froude number
of the upstream flow, where V1 is the average flow velocity.

12.5 Consider the flow as shown in Fig. 12-25

\
1. Draw two flow nets at two different scales to analyze
the pressure distribution at the entrance of the gallery
and against the gate. Figure 12-25
2. Calculate the integral of the horizontal and vertical
components of the pressure forces acting against the
jump. In these three cases it is assumed that the two end
gate. Compare the result of this total horizontal sum
water depths h 1 and h2 are each constant and that the fluid
with that obtained by applying the momentum
discharge per unit of width q is the same.
theorem.
By application of the momentum theorem and the
3. Is there any risk of cavitation? [Pressure head
Bernoulli equation to these three cases, determine the
< -32 ft (10m).] value of the external forces and the head losses (neglect the
4. Give the values of u and v and 8uj8x, 8uj8y, 8vj8x,
shearing stress at the wall).
8vj8y along OY for x = 0 at the entrance of the
gallery from A to B.
12.7 Find the value of the contraction coefficient in the
12.6 Consider the three following two-dimensional flows case of the circular orifice (called the Borda mouthpiece) as
as illustrated by Fig. 12-26. The first flow (1) is a sudden shown in Fig. 12-27. The contraction coefficient Cc is
enlargement under high pressure; the second flow (2) is a defined by the ratio of the smallest cross section of the jet to
gently diverging flow; and the third (3) is a hydraulic the cross section of the orifice.
 Chapter 12: The Momentum Theorem and Its Applications

151

.~----.....I:>.J>..I
a) ?~~----roM

Figure 12-27

12.10 Consider the case of a hydraulic jump created by an

abrupt drop h of the bottom of a channel. Demonstrate the
two following relationships between the upstream water
depth y 1 and the downstream water depth y 2
Figure 12-26

12.8 Consider a weir such as shown in Fig. 12-28. Deter-

mine the expression for z as a function of the head above or

~:- ~ 1 ~2 ~/Y1 [(;1 + ~y- e:YJ

the weir edge H. It will be assumed that the weir is aerated,
i.e., the atmospheric pressure is applied on the free surface.
The discharge per unit length is q = 0.5H(2gH) 1 12 and
V will be taken equal to 0.1(2gh) 112 What error is made depending upon the assumption for the value of the pressure
when we neglect angle rx of the falling water with the vertical? distribution on the vertical wall forming the abrupt drop.

12.9 A jet hits a plane perpendicularly. The discharge of the Figure 12-28
jet is Q = 2 ft 3 jsec (56630 cm 3 jsec) and the particle velocity
is V = 20 ft/sec (609 em/sec). The plane is moving at a
velocity U (U < V) in the direction of the jet. Calculate (1)
the total force exerted by the jet on the plane as a function of
the velocity U; (2) the power of the jet in horsepower; (3) the
power transmitted by the jet to the plane as a function of the
velocity U; (4) the efficiency defined as a ratio of these two
powers when U varies from 0 to V.
Do the same calculations assuming that the plane is
replaced by a bucket as shown on Fig. 12-29. I
 Part 2: Some Mathematical Treatments of the Basic Equations

152

Figure 12-29

12.11 Consider a horizontal convergent between two cross

sections A 1 = 2 ftl (1858 cm 2 ) and A 3 = 1 ft 2 (929 cm 2 ).
At section A 1 , the pressure p 1 = 12 psi (0.84 kgjcm 2 ) and
V1 = 6 ftjsec (182.8 cmjsec). The shearing force exerted by
. r = P.if V 2 , where f = 0.05 and V is the average
water IS
velocity as a function of the area of the cross section.
Determine the head loss and the total force exerted by the
convergent on its anchor as a function of the length of the Figure 12-30
convergent.
Determine the total force exerted on the anchor as a
function of the length of convergent in the case where the
convergent is bent by 90. 1. Calculate the discharge per linear foot of spillway
Now, neglecting friction force and taking a length of (or meter of spillway).
convergent of 10ft (304.8 em) determine the total force on 2. Draw the flow net by successive approximation.
the anchor in the straight and curved convergent in the 3. Determine the pressure distribution from the velocity
case in which vl is time dependent such that vl ft/sec = field and establish if there is any risk of cavitation.
6 sin (2n/T)t and T = 20 sec [or V cmjsec = 182.8 sin 4. This spillway is ended by a skijump (bucket) of
(2n/T)t]. 70 ft (21.34 m). radius as shown on Fig. 12-31.
Calculate the pressure distribution on the bucket
12.12 Consider the spillway defined by Fig. 12-30. The (without drawing any flow net) and integrate it in
coefficient of discharge C, defined by A = Ch(2gh) 112 , is a order to obtain the total force on the bucket.
function of h/hn such as 5. Determine the total force on the bucket by application
of the momentum theorem. Compare the results of 4
and 5.
h/h. 0.2 0.4 0.8
6. Calculate the distance D between the foot of the
c 0.394 0.425 0.470 0.490 dam and the location of impact of the jet.

h/h. 1.2 1.4 1.6 2 12.13 Establish, by choosing a number of simplifying

assumptions and by making use of the momentum theorem
c 0.504 0.518 0.532 0.552
and the Bernoulli equations, the set of equations giving the
distribution of discharges through manifolds with two
It will be assumed that hn = 8 ft (243 em) and h = 12 ft holes, three holes, four holes, ... , n holes of same cross
(365 em). section and subjected to the same pressure. The head
 Chapter 12: The Momentum Theorem and Its Applications

153

Figure 12-31

losses through the normal section of the main pipe will be

neglected.

12.14 A boat is equipped with a windmill. The axis of the

windmill is parallel to the axis of the boat. This windmill
generates the power used by the propeller. Can the boat
go upwind? If it can, establish the condition for this to be
done. Assume that efficiency of the system windmill-
propeller is unity, and that the drag of the boat is pro-
portional to the square of its velocity. Can the boat go
faster than the wind when she travels in the wind direction?
154
Chapter 13 13-1 General Concept of Boundary
Layer
13-1.1 Definition
The Boundary Layer, 13-1.1.1 As a viscous flow passes a solid boundary such
as a flatplate or a streamlined body, the influence of viscosity
Flow in Pipes, Drag, on the flow field is usually confined within a thin layer
near the boundary. Outside this layer, the effect of the
and Added Mass viscosity is vanishingly small, the fluid behaves like a perfect
fluid. This physical picture suggests that the entire flow
field can be divided into two domains, each of which can
be treated separately for the purpose of simplifying the
mathematical analysis, as shown in Fig. 13-1.
The first domain is called the boundary layer, which is a
thin layer right in the neighborhood of the boundary.
In this domain, the flow velocity is zero at the wall and
increases rapidly to the velocity corresponding to the free
stream velocity. In the boundary layer, the velocity distri-
bution is strongly influenced by the friction forces.
In the second domain, the influence of viscosity is small.
The friction forces can be neglected in comparison to the
inertia forces. Hence, the viscous terms in the Navier-
Stokes equations may be neglected. The fluid can be
assumed to be nonviscous and can be considered as
irrotational.
The pressure in the boundary layer, as will be shown later,
is approximately equal to the pressure at the limit of the
free stream.

Figure 13-1 Two flow domains.

FREE STREAM
 Chapter 13: The Boundary Layer, Flow in Pipes, Drag, and Added Mass

LAMINAR FLOW boundary layer. The corresponding decrease in flow rate is 155
TURBULENT FLOW
WITH LAMINAR SUBLAYER
L: 0 (U 0 - u)dy
TURBULENT FLOW

PERFECT FLUID
The displacement thickness 5* is the value by which the
wall will have to be shifted in order to give the same total
discharge if the fluid were frictionless. Consequently, 5*
is defined by the equality, as shown in Fig. 13-3,

(j*Uo = L: 0
(U 0 - u) dy

Figure 13-2 Relative velocity distributions as a I.e.,

function of the flow regime.
(j* = U1 Joo (U 0 -
1
u)dy ~- JiJ(U 0 - u)dy
o y=O Uo o

13-1.1.2 The larger the value of the Reynolds number,
the thinner the boundary layer.
Consequently, at very high Reynolds numbers, the
average flow motion with respect to time is very close to
that of a perfect fluid. This point has been explained
qualitatively in Section 8-1.1, and is further illustrated in
Fig. 13-2.
13-1.2.3 The momentum flux is reduced in the boundary
layer. As a measure of reduction of momentum flux, the
momentum thickness is then defined in a manner similar
to that for the displacement thickness. The reduction of the
momentum flux in the boundary layer is
L:/u(U 0 - u) dy

The momentum thickness 8 is then defined by

13-1.2 Thickness of Boundary Layer
13-1.2.1 The definition of thickness of the boundary pU68 = p L00
u(U 0 - u) dy
layer is to a certain extent arbitrary because the transition I.e.,

uz1 Joo uz JiJu(U 0

of velocity from zero to the ambient velocity takes place
asymptotically. Because the velocity increases very rapidly 8= u(U 0 - u)dy ~ 1 - u)dy
0 y=O 0 0
from the wall to the free system velocity, it is possible to
specify the thickness of the boundary layer beyond which
Figure 13-3 Displacement thickness (area ABC
the effects of the wall friction are rather small. For example,
equals area CDE).
the distance from the wall at which the velocity differs from
y E
the free stream velocity by 1 %is often used. Other quantities
used to describe the extent of the boundary layer are the
displacement thickness 5*, and the momentum thickness U (X ,y)
8. The significance of these quantities is defined in the
following.

13-1.2.2 Because of the existence of the friction forces

a certain amount of fluid is slowed down within th~
Part 2: Some Mathematical Treatments of the Basic Equations

value than the velocity u along the boundary layer. As a

13-2 Laminar Boundary Layer result, the terms that contain the velocity v in the second
equation are of much smaller value than are the terms in the
13-2.1 Steady Uniform Flow over a first equation, and they may therefore be neglected. Finally,
Flat Plate the second equation becomes
13-2.1.1 The Navier-Stokes equations in the case of a op*
two-dimensional steady motion are ay- =0.
u au + v au = - ~ op* + v(o 2 u + o2 u) This states that the pressure is hydrostatic along a per-
OX oy p OX ox 2 oy 2 pendicular to the plate and that p* depends only on x,
and can be determined from the nature of the flow in the
u OV + v OV = - ~ op* + v(o 2 v + ~) free stream. Because p* is a function of x only, one has the
ox oy p oy ox 2 oy2
equality
and the continuity equation is
op* dp*
ou + OV = 0 ox dx
ox ay
and the equations of motion become
As mentioned in the previous section, the velocity
varies rapidly along the y axis. It is zero at the wall and ou ou dp* o2 u
u - + v - = - - + v -2
reaches the free stream velocity at a distance of the order of ox oy dx oy
the thickness of the boundary layer J. In contrast, the which is often called the boundary-layer equation. In
velocity varies very slowly along the plate (see Fig. 13-4). addition to the above equation, the boundary conditions
Therefore, all derivatives in the y direction must be much
larger than the derivatives in the x direction (see Sections u= v= 0 at y= 0
4-5.2.4 and 5-4.2). Consequently, in the first equation
u = U0 when y-+oo
ou
v-~v -
2 ou
2
and the continuity equation determine the flow field near
ox2 oy2 the fiat plate.
and the term v(o 2 ujox 2 ) may be neglected. Furthermore, In the particular case in which U 0 is constant, i.e., in the
the velocity v across the boundary layer is of much smaller case of a steady uniform flow over a fiat plate with zero

u.
u.

Figure 13-4
Schematic drawinq of the boundary layer.
 Chapter 13: The Boundary Layer, Flow in Pipes, Drag, and Added Mass

incidences, as a consequence of the Bernoulli equation 13-2.1.3 Let us now consider the case of steady flow 157
applied in the second domain, one has (ap*jax) 0. = over a flat plate where ap* ;ax = 0, and let us introduce the
stream function 1/J(x,y) such that u = aljJ ;ay, v = - (aljJ ;ax)
13-2.1.2 As shown in Fig. 13-4, the momentum flux whereby the boundary-layer equation becomes
per unit width through OM is equal to pU6f>. The momen-
tum flux through AB is smaller. However, because u < U 0 aljJ a2 aljJ a2
t/l a3 t/J t/J
------=V-
in the boundary layer, the definition of shows that < f>.
e e ay axay ax ay 2 al
Hence, the difference of momentum is also of order pU6f>.
In contrast, the total friction force per unit width between which is a third-order nonlinear differential equation. From
between the sections 0 M and AB is the above definition it is seem that t/1 is equal to Jb u dy. If
u = U 0 <1>(ry), and y = ry(vxjU 0 ) 112 are substituted in this
equality the result is

t/1 = f(ry)(U 0 vx) 112

By definition of boundary-layer thickness, aujay ,..,., U 0 /f>.
Since the difference of momentum flux between sections
0 M and AB is caused by the friction force along 0 B, one
may equate the difference of momentum flux with the
friction force. Consequently, one has
Note that y is considered as a function of IJ.
The boundary-layer equation presented above can now
be transformed into an ordinary differential equation as
follows:
aljJ aljJ a11 1
u = ay = arr ay = f'(ry) J (U 0 vx) 112 = U 0 f'(ry)

or 112
v= - -aljJ = _o) (ryf' - f)
1- (vU
b ""' (~)1/2 = (~)1/2 ax 2 X
pUo Uo
au = a21jJ = u o(u 0)1/2 f"(ry)
This equation is valid only in the case of uniform pressure ay ai vx
distribution along the plate. If it is assumed that the velocity
profiles at all distances x from the leading edge are similar, au= a21jJ =- U 0 11 f"(ry)
ax axay 2x
the velocity profiles u(y) for varying distances x can be
made identical by choosing the proper scale factors for
u(y) and y. The free stream velocity U 0 and the boundary
a31jJ = u
al
o(uvxo) f"'(ry)
layer thickness f> are these scale factors. Hence, one has
the similarity relationship Substituting these into the boundary-layer equation, one
obtains the Blasius equation 2f"' + ff" = 0, and the
u boundary conditions f = f' = 0 at 11 = 0, and f' = 1
- = <l>(ry)
Uo atry=oo.
defining <l>(ry), where 11 = yjf>, i.e.,
13-2.1.4 The general solution of the Blasius equation
U 0 ) 1/ 2
( -
cannot be given in a closed form. However, the solution
IJ = y
vx can be obtained through power-series expansion. The
Part 2: Some Mathematical Treatments of the Basic Equations

power-series expansion near 17 = 0 is assumed to be of f"(O) = 0.332. Therefore, the shear stress at the wall is

(oyau) y=o = J.1Uo (uvx

the form of
1]2 1'/3 0 ) 1; 2 , 0.332 2
f(rJ) = A0 + A 1 17 + A2 2 ! + A3 3 ! + ro(x) = f.1 f (0) = (RJ1!2 pUo

where Rx = U 0 xjv is the Reynolds number based on the

where An are constants. From the boundary conditions
distance x from the leading edge of the plate.
f = f' = 0 at 1J = 0, one obtains A 0 = A 1 = 0. The resistance force on one side of the plate over length l
Substituting the power series with A 0 = A 1 = 0 into the
per unit width is equal to
Blasius equation, one obtains
2 1'/2
_ ( d _ 0.664 pU6l
2A 3 + 2A 4 1] + (A 2 + 2As) 2 ! D- Jo'o X - (Uol/v)112

1'/3 and the resistance coefficient C I is

+ (4A 2A 3 + 2A 6 )3T + = 0
D 1.328
This must be equal to zero for any value of 1], which can be CI = !plU~ = (R,)112
verified only if all the coefficients of each term are equal to
where R 1 = U 0 ljv is the Reynolds number based on the
zero. Consequently, one has
length of the plate.
A3 = A4 = A6 = A7 = A9 = = 0
As= -!ALAs= -VA2As = 1iA~; Figure 13-5 Solution of the Blasius equation.
1.0 r------.----.---.---,,----,---,:::::oo"'
Hence, all the coefficients can be expressed as a function
of A 2 The constant A 2 can be determined from the second
boundary condition: y --HfJ, u = U 0 ; or 17 --HfJ, f' = 1.
Once A 2 is obtained, f(rJ) can be calculated. The solution
f(rJ) together with f'(rJ) and f"(rJ) are plotted in Fig. 13-5.
This gives the Blasius solution of the laminar boundary-
layer equations.
f I
13-2.1.5 The boundary-layer thickness for the steady
uniform flow over a flat plate as defined in Section 13-1.2.1 f II

(that is, the distance from the wall at which u = 0.99 U 0 )

can be obtained from Fig. 13-5, where 17 ::::: 5.0, i.e.,
b
11 = (vxI U 0 )112 ::::: 5.0
Hence, the boundary-layer thickness b becomes
X
b ::::: 5.0 (xU o/v)112
2 3
13-2.1.6 From the numerical calculation in Section ru:
13-2.1.4, or the graph presented in Fig. 13-4, one obtains 7J=yV~
 Chapter 13: The Boundary Layer, Flow in Pipes, Drag, and Added Mass

13-2.2.2 The discharge through AD is 159

13-2.2 Momentum Integral Equation for
Boundary Layer J:u dy lx;x,
13-2.2.1 As demonstrated in the previous simple example and the discharge through CB is
of a laminar boundary layer on a semiinfinite flat plate
at zero incidence, the calculations are cumbersome and f ~u dy I +dx = J~u dy I + !!_ (f~u dy) dx
0 X;Xl 0 X;Xl dx 0
time consuming. It is desirable for practical use to find
some approximate method to evaluate the necessary The net outflow over the vertical control surface is equal
quantities. In particular, the case in which (8p*j8x) can no to the difference of these discharges:
longer be considered as zero, such as in flow past a wedge,
can be analyzed by the momentum integral method
developed by von Karman.
Consider an element ABCD of a two-dimensional steady This amount of flow must be supplied through the top
flow as shown in Fig. 13-6. The momentum integral method boundary for the sake of continuity.
consists of applying the momentum theorem to this element,
i.e., equating the variation of momentum flux among the 13-2.2.3 Similarly the net x momentum outflow through
boundaries AD, BC, and CD, to the applied forces. The a vertical control surface is equal to
applied forces consist of the pressure force acting on the
boundaries and the shear stress on the wall. Each of these
will be considered separately. The equality will give the
momentum integral equation. Of course, in the case in On the other hand, because the velocity at the limit of the
which the flow is unsteady, the additional term resulting boundary layer is U 0 , the x momentum inflow over the top
from unsteady flow also has to be considered in the equation. is equal to the mass flow pU 0 times the fluid discharge

u.- u. u.
C LIMIT OF BOUNDARY
----~~~~~--~LA~YEER

Figure 13-6
Diagram for the momentum integral method.

8 ------------~---x
oL-----~--~------------~

I
t +d
 Part 2: Some Mathematical Treatments of the Basic Equations

160 through the control surface. This fluid discharge was
shown in the previous section to be
Therefore, the x momentum inflow over the top boundary
DC is
13-2.2.5 Now that all the terms have been established,
it is possible to write the momentum integral equation
for the element ABCD. One obtains the momentum integral
by equating the momentum flux (in the x direction) and
the net pressure force to the shear force on the boundary,
that is,

and the total variation of x momentum flux through

ABCD is By dividing the above equation by dx and rearranging
the terms, one obtains

d
To= dx [fb/u(U 0 - u) dy J+ dU0fb/(U 0 -
dx u) dy
13-2.2.4 The pressure forces acting on the limit of the
element ABCD in the OX direction are now considered. Introducing the displacement thickness, b* and the
The pressure forces on AD are pb, and on CB they are momentum thickness (), the momentum integral can be
written

To=!!.__ (U68) + b*Uo dUo

p dx dx
Since the variation of boundary-layer thickness with
distance is small, (dbjdx) dx may be neglected, but (dpjdx) dx Because no assumption is being made as to the nature of the
may not be small. flow, this method is applicable to laminar as well as
The pressure force on DC acting in the OX direction turbulent flows. However, in turbulent flow, the velocity
is also neglected because db/dx is small. should be considered to be the mean value with respect to
Finally, the net pressure force remains: - b(dpjdx) dx, time.
This last term may also be expressed in terms of the velocity
U 0 as follows: 13-2.2.6 The momentum integral method is now applied
From the Bernoulli equation applied to the irrotational to the simple case, already investigated, of a steady uniform
flow outside the boundary layer, one has laminar flow over a flat plate.
In this case, U 0 is constant. Therefore, the momentum
dp=-pU 0 dU 0 integral becomes
dx dx To d()
Hence, the net pressure force is pU6 dx
First, one assumes a velocity profile such as
 Chapter 13: The Boundary Layer, Flow in Pipes, Drag, and Added Mass

where a0 , at> a 2 are constants that can be determined from Both the boundary-layer thickness and normalized shear 161
boundary conditions. For example stress are close to the exact value obtained in Sections
13-2.1.5 and 13-2.1.6.
u=O at y=O implies a0 = 0
u = U0 at y=b implies a 1b + a2 b2 = U0 13-2.2.7 For flows with pressure gradient, Pohlhausen
suggested using a fourth-order polynomial
du ~ 0
at y=b implies a 1 +2a 2 b=0
dy- u
U = ao + a 1 1J + a2 1J 2 + a 3 1J 3 + a4 1J 4
Therefore
and adding to the boundary conditions used above, the
Uo
a1 = 2-y conditions
o2 u 1 dp dUo
Uo v - = - - = - U0 - - at
a2 = - -p- oi p dx dx

and
and the velocity profile is
at y=b

13-2.3 Uniform Unsteady Flow over an

One can find the momentum thickness fJ to be

J
Infinite Flat Plate
f) = s: ~ 0 ( 1- ~ dy = 125 b
13-2.3.1 Because the plate is of infinite length, the
derivatives with respect to x should be zero. That 1s,
and oujox = 0. From the continuity equation
To au I 2vU 0
au+ ov = 0
---;; = v ay y=o b ax ay
Substituting these equalities into momentum integral it follows that ov/oy = 0. Hence, v is identical to zero
equation, one has because it is zero at the boundary. Furthermore, the
15v = b db pressure p* is constant everywhere, because of the infinite
U0 dx
fluid field. Finally, the Navier-Stokes equation becomes

Integrating and using the boundary condition x = 0,

au 2u
-=V-
o
b = 0, one obtains Ot 2 oy

b = ( vx)l/2
5.5-
This equation is linear. A number of exact solutions can
be found as functions of the boundary conditions. If the
Uo
fluid is moving at a speed U 0 (t) and the plate is fixed, the
and the normalized shear stress is equal to boundary conditions are

'o v au I v 2U 0 0.366 u=O at y=O

pU6 = U6 oy y=O U6 . -b- (Rx)t/2 u = U 0 (t) when y -HXJ
 Part 2: Some Mathematical Treatments of the Basic Equations

162 If the fluid at infinity is fixed and the plate is moving at or

velocity U 0 (t), the boundary conditions are y
u = U 0 erfc (2 vt) 112
u = V 0 (t) at
u = 0 when The velocity distribution is presented in Fig. 13-7.
It is clear that the velocity profile for different times are
13-2.3.2 The case of an impulsive motion of an infinite similar. They can be reduced to one curve by using the
flat plate is given as an example. In that case, u = 0 for all dimensionless variables ujU 0 and IJ = y/(2vt) 112
y when t :::; 0, and u = U 0 at y = 0 when t > 0, and u = 0
13-2.3.3 As an infinite flat plate oscillates parallel to
when y -HfJ.
itself, the governing equation of motion is the same as the
The partial differential equation can be transformed into
impulsive motion of an infinite flat plate, that is,
an ordinary differential equation by introducing the
dimensionless variable 2au au
y
-=V-
at
2 ay
IJ = (2vt) 112 The boundary conditions may be given by

After performing the required differentiations and sub- u(O, t) = U 0 cos kt

stituting into the equation, one obtains u(oo, t) = 0

d2 u du as the plate oscillates periodically.

d1]2 + IJ dl] = 0
Figure 13-7 Velocity distribution near the infinite
Integrating with respect to IJ, one gets fiat plate under impulsive motion.

u(y,t) = c1 0J y/(2vt)l/2
e~~ 212 dl] + c2
1.6

where c2 can be determined by the boundary condition

u = U 0 , for 1J = 0. This gives C 2 = U 0 . The constant C 1 1.2

is determined from the initial condition u = 0 at t = 0

\
\
(IJ = y/(2vt) 112 = oo ). Substituting into the above equation,
one has
o.e
c1 Lx>~2j2 dl] + Uo = 0 \
The above integral has a value of (n/2) 112 ; hence C 1 = 0.4
~
-(2/n)112Uo.
Substituting C 1 , C 2 into the equation, one obtains the
~ I'..
velocity distribution
"""' ~
u = U{ 1 - (2/n)1/2 J:/(2vt)'!2 e~~2 dl] J 00 0.2 0.4
u
0.6 0.8 1.0

Uo
 Chapter 13: The Boundary Layer, Flow in Pipes, Drag, and Added Mass

4
ljl\
Figure 13-8
Velocity distribution near an
3
kt'
3,.
2 ~ v,
v !/ '\ "(
J 1\
kt 'f

~v
oscillating infinite flat plate. 2
I
I

~
""'
kt 'O,OR2,.
kt ' " \

__.... v
v """ ~ ]/
[7 ~
1'-----t--..
0
- 1.0 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1.0
.....!!....
uo

The solution of this equation is
The velocity profile, u(y,t) has the form of damped harmonic
oscillation, with an amplitude of U 0 exp [ -(k/2v) 1 12 y].
The amplitude decreases exponentially from the plate.
The velocity profiles for several instants of time are plotted
in Fig. 13-8.
A similar solution applies in the case where the fluid is
moving at a velocity u = U 0 cos kt, and the plate is fixed.
Such solution is of particular interest for studying the
motion in the boundary layer of a periodic gravity wave
and the wave damping by bottom friction.
13-3 Turbulent Boundary Layer
13-3.1 General Description
13-3.1.1 For a laminar boundary layer, it has been seen
that its thickness increases with the distance x from the
edge of the plate. As this boundary-layer thickness increases,
the flow has a tendency to become turbulent. The criteria
of the transition from laminar to turbulent are usually
based on the Reynolds number U 0 x/v, in which U 0 is the
free-stream velocity, x is the distance from the edge of the
plate, and v is the kinematic viscosity.
The value of Reynolds number (or the location x) at
which the boundary layer becomes turbulent depends
somewhat on the turbulence level of the free stream; it
ranges from 10 5 to 10 6 . The shear stress acting on the
boundary is much larger in the turbulent boundary layer
than in the laminar boundary layer; therefore, the deter-
mination of the location of this transition is not only of
theoretical interest but has practical uses.
After transition, the main part ofthe flow in the boundary
layer is turbulent. However, immediately adjacent to the
wall, the turbulent fluctuations are suppressed by the
presence of the wall. The flow field in this region can be
divided into three domains-the laminar sublayer, the
turbulent boundary layer, and the free stream (see Fig. 13-9).
If the boundary is rough, the laminar sublayer may be
destroyed by the presence of the roughness elements.
A detailed discussion of this is given in Section 13-4.3.
 y
164

--r
FREE STREAM

T
u ( y)

TURBULENT BOUNDARY
y LAYER

!
TRANSITION

/(
o) SMOOTH WALL

Figure 13-9
Turbulent velocity distributions

t-
FREE STREAM
near the wall.

T LAYER
BOUNDARY

b) ROUGH WALL {LAMINAR SUBLAYER MAY

k 5 u.
EXIST PROVIDED -- <5
!I
 Chapter 13: The Boundary Layer, Flow in Pipes, Drag, and Added Mass
Figure 13-10
Schematic representation of the laminar and turbulent
boundary layer.
13-3.1.2 The velocity distribution of a laminar sublayer
is determined by the viscous force, that is, T = pv(dujdy).
Because this layer is very thin, it is reasonable to assume
that T is constant within this layer and equal to the shear
stress at the wall To. After integration, the equation becomes
Toy
u= --
p v
If, by definition,
u* = (~ y/2
then one finally obtains
u* is called the shear velocity.
The name laminar sublayer does not mean that the flow
in this region is entirely laminar. Strong eddies generated
in the turbulent flow often break through this thin layer
and form turbulent spots in the sublayer. Therefore, to
avoid confusion, the name viscous sublayer is sometimes
used.
13-3.1.3 In the case of a turbulent boundary layer, the
effect of turbulent fluctuation creates a large turbulent
shear stress, whereas the effect of viscous shear is very
small. Therefore, the velocity distribution is determined
LAMINAR ~RANSITION TURBULENT
by the effect of turbulent shear stress which results in a
logarithmic velocity distribution, as discussed in Section
8-3.5.
13-3.1.4 The effect of the boundary shear stress in the
case of free stream flow is small. Therefore, the flow field
can be determined by considering that the flow is non-
viscous.
However, there are no sharp boundaries between each
region, and the concept of each domain is to some extent
qualitative. A schematic drawing of the three domains
of flow and the flow pattern before the formation of the
turbulent boundary layer is given in Fig. 13-10.
13-3.2 Resistance and Boundary-Layer
Growth on a Flat Plate
Owing to the complicated flow conditions in the
turbulent boundary layer, the exact solution of the equation
of motion is not possible. One mathematical method
available at present consists of determining the character-
istics of the turbulent boundary layers by application of the
momentum integral method described in Section 13-2.2.
The momentum integral method is used in the turbulent
boundary layer to evaluate the variations with distance
of the thickness of this boundary layer and the boundary
shear stress. The use of this integral method involves the
assumptions of a velocity profile at one location and similar
profiles along the boundary. In the turbulent boundary
 Part 2: Some Mathematical Treatments of the Basic Equations

166 layer, it is rather difficult to assume a velocity profile the boundary layer starts to become turbulent, to a given
with sufficient accuracy because of the complicated flow point x measured from the leading edge of the plate.
field. Therefore, some experimental results have to be used The result is
as a base in order to get a good prediction of the variations
of thickness of the boundary layer and the resistance with
distance. Based on the experimental results the velocity
D514 - D614 = 0.29(;J 114
(x - x 0 )

profile can be represented by If one assumes that the boundary layer becomes tur-

8.74(u*y) 117
bulent at the edge of the plate, the initial value above can
!I_= be replaced by D = 0 at x = 0. The above equation gives
u* v directly
where u* = (T 0 /p) 112
is the shear velocity. Because the
u )-1;5
velocity at y = Dis equal to U 0 , then D = 0.37(-;- x 415

Uo = 8.74(Du*)1f7 This indicates that the boundary-layer thickness increases

u* v with the power x 415 , whereas in the laminar boundary layer,
or T 0 = 0.0225 pU5(v/U oW 14 . From the first two equations, the thickness increases with x 112 .
one can immediately derive The resistance force per unit width of length l is

R = {To dx = pU68(/)

From the definition of the momentum thickness, one has = / 2 pU6D = 0.036 pU6 (-;-
u )-1;5 1
415
successively
and the resistance coefficient C1 is
R (Uol)-115
cf = 1-----z/ =
zPUo
0.072 -
v
In the range of Reynolds number 5 x 10 5 < Re < 10 7 , the
last equation gives very good agreement with experimental
On substituting momentum thickness (} and the wall shear results.
stress To into the momentum integral equation, The value of the resistance coefficient for the turbulent
boundary layer as well as the laminar boundary layer are
To d(}
plotted in Fig. 13-11.
U0 = constant
pU6 dx
one obtains 13-4 Flow in Pipes
0.0225(-v-)
U D 0
114
= 272 db
dx
13-4.1 Steady Laminar Flow in Pipes
Laminar flow in pipes rarely occurs in practice. However,
This is the differential equation for D. This equation is a rather full discussion is given because the simple and
integrated from the initial value D = D0 at x = x 0 , where rational analysis is of some help in the understanding of
 Chapter 13: The Boundary Layer, Flow in Pipes, Drag, and Added Mass

10
-2
After integration, the velocity profile becomes

-- - -- Lip r 2
u =--+c

-
t- TURBULENT 2/f.l 2
~ TR~
~ - --- ----- t- The constant c can be determined from the boundary
condition u = 0 at r = r 0 . Therefore, the velocity distri-
'-
.......... bution in the pipe is
t'- ~INiR
-3
10 Lip 2 2
10 6 106 6 107
u = - (r 0 - r)
R = Uo/ 4/f.l
e 1/
which has the form of a symmetrical paraboloid. By integra-
Figure 13-11 Resistance on a smooth fiat plate.
tion, the average velocity V

the turbulent flow where conditions are so complicated that

V = -Q2 = - 12
nr 0 nr 0
1'"0
u2nr dr L1pr6
=~81f.1
a complete theoretical treatment is not possible. Rearranging the above equation, one obtains
The flow conditions can be determined directly from the
application of the Navier-Stokes equation. However, it is Lip- 64 f.ll v2
much simpler to derive the equation of motion directly pg- VpD 2 2g
from a consideration of force equilibrium. Consider a
where D = 2r 0 .
cylindrical element of length l and radius r (Fig. 13-12).
This equation can be compared with the Darcy-
The equilibrium of the shear stress T and the pressure drop
Weisbach equation for head loss, obtained by dimensional
Lip is
analysis, that is
L1pnr 2 = 2nnl
L1H = Lip = f _!_ v2
It is assumed that Lip is a constant across a pipe section, pg D 2g
a result which can be derived from the Navier-Stokes
equation. For the laminar flow, the shear stress is simply, where L1H is the head loss and f is the friction factor. One
T = f.l(dujdr). Substituting into the above equation, one
obtains immediately the friction factor f for the laminar
obtains flow in a circular pipe

Lip f = 64J.1 = 64
du = 21 f.1 r dr VDp R.

13-4.2 Turbulent Velocity Distributions

and Resistance Law for Smooth Pipes
Figure 13-12 Force equilibrium for a cylindrical element.
13-4.2.1 As the flow in the pipe becomes turbulent, the
analytical determination of the velocity distribution is
not possible. As in the case of turbulent boundary layer,
the velocity profile must be determined based on logical
assumptions and experimental verifications. One of the
 Part 2: Some Mathematical Treatments of the Basic Equations

168 best-known assumptions in regard to the velocity distri-

-- -
LAMINAR
bution near the wall is based on the assumption that the FLOW
velocity u at a distance y from the wall depends on the
tangential stress r 0 and on the viscosity J1 and density p. SMALL Re..... '\
-+-+-
Therefore, one may write in the most general case that
F(r 0 ,u,y,p,p) = 0 (It is understood that u means the average
velocity with respect to time and should actually be written
ii. The bar will be omitted in the following sections for the ---
sake of simplicity.) Based on the dimensional analysis,
one obtains the dimensionless form a) SMOOTH WALL

u = [('oY/P)l/2]
(ro/P) 112 f v

which is similar to the case of flow over a flat plate (see

Section 13-3.2). The functional relationship in the laminar
sublayer has been derived in Section 13-3.1 to be
/
/
LARGE
ROUGHNESS

where u* = (r 0 /p) 112 . Outside the laminar sublayer and a

transitional layer the turbulent stress dominates and the b) ROUGH WALL
velocity profile follows logarithmic law, which has been Figure 13-13 Typical velocity distribution in pipes.
derived in Section 8-3.5 as

Many experiments have been performed for measuring

the velocity distribution in circular pipes. Typical velocity
profiles are shown on Fig. 13-13a. The results are also
In writing the dimensionless form, the velocity distribution
presented in terms of the dimensionless variables uju*
reads
and u*y/v in Fig. 13-14. It indicates that at low values of
u*yjv, (u*y/v < 5) the velocity follows the linear relation-
ship
where
1 u*
C 2 =--In-+ C 1
k v For values of u*yjv > 30, the experimental curve follows
the logarithmic law which can be approximated by the
The value of C 2 and the range of validity of these two equation
equations which describes the velocity distribution in
laminar sublayer and the turbulent flow have to be deter- -u = 5.75log 10 -yu* + 5.5
mined experimentally. u* v
 Chapter 13: The Boundary Layer, Flow in Pipes, Drag, and Added Mass

v v
30 !69

TURBULENT

/
20
v
//
Figure 13-14
Universal velocity distribution law for smooth pipes. 10
21!' l-/
TRA'SOTION

MINAR
I
0
2 3 4 5
10 10 10 10 10

';y, NORMALIZED DISTANCE FROM THE WALL

However, in the range u* yjv between 5 to 30, 'Yhere both After integration, one obtains
turbulent and viscous effects are of equal importance, the
velocity profile deviates from both of the above equations.
Mathematical analysis fails to give correct prediction. This
region is usually called a buffer region or a transition region.
On the other hand, it has been seen that the Darcy-
Weisbach equation gives the value of the head loss I!.H
13-4.2.2 Usually the thickness of the laminar sublayer as function of the friction coefficient f and the average
and the layer of transition is very small in comparison with velocity U as follows:
the size of the pipe. Therefore, in computing the average
velocity, one could just use the logarithmic velocity distri- I U2
bution without introducing any significant error. The I!.H =f--
D 2g
average velocity may then be obtained by substituting
where I and D are the length and diameter of the pipe,
u yu*
~ = 5.75log 10 - + 5.5 respectively. Also, pg 11H(nD 2 /4) is the difference of pressure
u* v forces acting on two cross sections separated by a distance
I. This force is balanced by the shearing force To nDI.
into the following equation: Equating these two forces yields

1 fro I!.H =~To_!_

U= - 2- 2nru dr
ron o gpD

2u* fro r(5 .75 1oglo u*(ro - r)

-- -2 + 55)
. dr Inserting u* = (T 0 /p) 112 and eliminating I!.H by considering
ro o v this above equality with the Darcy-Weisbach equation
 Part 2: Some Mathematical Treatments of the Basic Equations

yields where ks is the sand size. (If the sand is not closely packed
(8)f
170
or nonuniform, one should also take into account the
-U =
1 2
- 1 = u r
5.75 log 10 ____>i:___Q + 1.75 concentration distribution and shape of the roughness
u* v
elements.) In the case in which the wall is not completely

(!)
Since
rough, then an additional dimensionless parameter k 5 u*jv
u*r-0 -
- _ 1
--
112
UD
- should also be included. Therefore, the general function
v 2 8 v should read
The above equation can be further written as

f 1112 -- 2.04log 10 Ref 1/2 - 0.91

This general functional relationship has been determined
where Re = U Djv. By modifying the constant slightly by experiments and can be approximated by the equation
to agree with the results obtained from experiments, that u y
is to change the constants in the above equations 2.04 - = 5.75 log 10 -k +B
and 0.91 to 2.0 and 0.8, respectively, one has u* s

where B depends on the "shear Reynolds number," ksu*jv.

! 1112 -- 2.0 log 10 Ref 1/2 - 0.8 The value of B obtained by experiments is shown in Fig.
13-15. As indicated in this figure, the value of B characterizes
which is Prandtl's universal law of friction for smooth pipes. three regimes:
1. Hydraulically smooth regime previously considered
13-4.3 Effect of Roughness
in Section 13-4-2.1, where
13-4.3.1 The effect of the roughness element on the
flow depends on the thickness of the laminar sublayer. 0 < ks/1* < 5
If the laminar sublayer is so thick that it covers the rough- - v -
ness, the roughness has no effect. The surface then can be
considered to be hydrodynamically smooth. If the size of
the roughness elements is large compared with the laminar Figure 13-15 Roughness function B in terms of shear Reynolds
sublayer, the effect of viscosity becomes small and no number k 5 u*jv.
longer enters explicitly into the picture. The surface is
v
II
then considered to be completely rough. In this case the
shear stress depends only on the roughness, the specific I
density p, and the velocity u at some distance y from the /.-' ~ ........_

/
9
wall. Some typical velocity profiles obtained in rough
pipes are presented in Fig. 13-13b. Following the same 8
procedure for the flow in smooth pipes, one could establish
7
a dimensionless functional relationship for the closely SMOOTH- - TRANSITION COMPLETELY ROUGH -
packed uniform sand roughness elements in completely
rough regime:
0.8
I
1.4 2.0
I
2.6 3.2

log
 Chapter 13: The Boundary Layer, Flow in Pipes, Drag and Added Mass

In this regime, the size of roughness is so small it
is covered by the laminar sublayer.
2. Transition regime:
5 S ksfl* S 70
v
Some of the roughness elements extend outside the
laminar sublayer and contribute some resistance
through form drag.
3. Completely rough regime:
ksfl* > 70
v
All the roughness elements are exposed outside the
laminar sublayer or one may say that the laminar
sublayer has been destroyed completely by the
roughness elements. The turbulent action extends
all the way to the rough wall. Further increase of
shear and the Reynolds number does not bring any
change of flow patterns. Therefore, B remains
independent of shear Reynolds number.
13-4.3.2 The resistance coefficient in the completely
rough regime can be evaluated the same way as in the case
Figure 13-16 Resistance formula for pipes roughed with
Nikuradse's sand roughness.
lOr---------~----------~---------,
of smooth pipe. The final form of the resistance equation
reads
1 r0
! 112 = 2.0 log 10 ks + 1.74
Experiments were performed first by Nikuradse, who
used closely packed sand grain roughness elements and
obtained the resistance diagram shown in Fig. 13-16.
The velocity distribution and resistance formula dis-
cussed are based on the closely packed sand grain roughness
used by Nikuradse. In this case, ks is the actual sand size.
However, if a different type of sand is used or sand particles
are not packed closely, the resistance offered to the flow
will be different. Therefore, sand size alone is not enough
to describe the velocity distribution and resistance.
13-5 Drag on Immersed Bodies
13-5.1 Draq on a Body in Steady Flow
13-5.1.1 It has been seen that the total force exerted by a
current on a cylinder is zero (see Section 11-4.4.3). This
result is general, for the case of a perfect fluid, i.e., the total
force exerted on a body by a perfect fluid without circulation
of velocity is always zero. This is called the paradox of
D' Alembert.
For the case in which a circulation is introduced to the
fluid (see Sections 11-4.2 and 11-4.3) a force perpendicular
to the incident velocity is exerted on the body. It can be

'o
-=15 demonstrated that this force is proportional to the velocity
ks
V of the fluid and the strength of the circulation. It is this
force that causes the lift of an aerofoil.
1001
The problem now under study is that of a real fluid in
which case a boundary layer develops along the body and
induces a drag. This drag is caused by the shearing force
acting on the body and to the wake. This leads us to discuss
the problem of boundary-layer separation.

1.0 L------------L.----------..l.---------__J
13-5.1.2 The flow field near a flat plate in parallel flow
103 10 4 10~ 106 and at zero incidence is quite simple because the pressure
R
e :~
v in the entire flow field remains constant. In the case of flow
Part 2: Some Mathematical Treatments of the Basic Equations
LIMIT OF BOUNDARY
LINE OF FLOW SEPARAT ION
about a blunt body such as a cylinder, the pressure along
the surface of the body, which is impressed on the boundary
from the external flow, is not constant. As shown in Fig.
13-17 the fluid particles are accelerated from A to B and
decelerated from B to C. Hence, the pressure decreases
from A to Band then increases from B to C. This can be
seen by application of the Bernoulli equation. Since the
fluid is viscous, a certain amount of kinetic energy is lost
by the friction within the thin boundary layer as the fluid
particles move along the boundary. The remaining energy
may be too small to overcome the increasing pressure
toward the point C. As a result of this, the fluid particles
being influenced by the external pressure may move in the
reverse direction and cause flow separation behind the
body at point S. The flow field behind the separation is
very irregular and is characterized by large turbulent
eddies. This region is usually called turbulent wake,
although the wake may also be laminar when the Reynolds
number is small ( < 40 in the case of a cylinder).
Because of the existence of the wake, the flow field
changes radically as compared with that in frictionless
flow. The main flow which separates from either side of
the boundary does not meet right behind the body. It leaves
a "dead zone" in which the pressure remains close to its
value at the separation point, which is always less than the
pressure at the forward stagnation point. Therefore, a
large net force will act on the body resulting from the
pressure difference. This force is called form drag.
Figure 13-17
Boundary-layer separation.
13-5.1.3 In principle, the total drag exerted on a sphere
moving with constant velocity in the infinite flow field
is the sum of the friction drag (or shear drag) and the form
drag. If the velocity is low enough (R. < 1), the inertia terms
in the Navier-Stokes equations may be neglected. The
drag can be obtained analytically and is given by Stokes'
law
F = 3npvVD
where V is the relative velocity of the body with respect to
the water and D the diameter of the sphere. The drag
coefficient Cv, which is defined from the equation
CvApV 2
F = --,-2- -
is then equal to 24/R., where A is the cross-sectional area
of the sphere, R. is the Reynolds number VDj v.
As the Reynolds number R. increases, the flow separates
from the surface of the sphere, beginning at the rear stagna-
tion point, where the adverse pressure gradient is greatest.
As the flow separates from the boundary, the form drag,
which is a function of the area of separation and the square
of velocity, becomes important. The drag coefficient C v
will deviate from the line C v = 24/R., and start to level off.
Figure 13-18 indicates the variation of the drag coefficient
C v with Reynolds number R.
As the Reynolds number reaches 2 x 10 3 , the drag
coefficient becomes almost constant. However, in the
 Chapter 13: The Boundary Layer, Flow in Pipes, Drag, and Added Mass
400
""~
100
Figure 13-18
Drag coefficient for sphere.
10
STOKES " ~
I
range of Reynolds number 2 x 10 5 to 3 x lOS, the drag
coefficient is suddenly reduced. The reason for this lies
in the transition of boundary layer from laminar to tur-
bulent. This transition brings a violent mixing in the
boundary layer. As a result, the fluid particles near the
boundary gain additional kinetic energy, which enables
them to withstand the adverse pressure gradient better
and to move the separation point somewhat downstream,
as illustrated in Fig. 13-19. This results in a sudden de-
creasing of the drag coefficient near the Reynolds number
3 x 10 5 as shown in Fig. 13-18. As the transition that occurs
depends on the roughness of the sphere, and also slightly
on the turbulence level in the free stream, the drag coeffi-
cient near this critical region is not a unique function of
Reynolds number.
13-5.1.4 As shown in Fig. 13-20, the relationship between
drag coefficient and the Reynolds number for a circular
cylinder with axis normal to the direction of motion is,
in general, similar to that for a sphere. However, rather
peculiar phenomena which are not ordinarily found in the
flow around a sphere can be observed in flow around a
cylinder. In the range of Reynolds number between 40 and
173
THEORY~
~
-
I~
~
5000, one can see a regular pattern of vortices which move
alternately clockwise and counterclockwise downstream,
as shown in Fig. 13-21. This is known as the Karman
vortex street. The vortex street moves with a velocity V.,,
which is somewhat smaller than the free stream velocity
U 0 Von Karman found that the vortex street is unstable
except at the spacing h/1 = 0.281, and that the drag
experienced by the cylinder depends on the width of the
vortices hand on the velocity ratio V.,/U 0 :
F = pU 02 h[ v., 1.12( Uo
2.83 Uo- v., ) 2
]
Because the vortex developed behind the cylinder is
unsymmetrical, a time-dependent circulation of velocity
is induced around the cylinder. The cylinder will experience
a side push that continually reverses its direction. Therefore
the cylinder may tend to oscillate from one side to the other,
particularly if its natural frequency of oscillation is in
resonance with the frequency of the vortex shedding.
The shedding frequencies k in the Karman vortex street
behind a circular cylinder have been measured. Those
measurements indicate that the dimensionless frequency
 Part 2: Some Mathematical Treatments of the Basic Equations
174
Figure 13-19 Changing location of separation and
pressure distribution as a result of
boundary-layer transition.
known at the Strouhal number S = kD/U 0 depends
uniquely on the Reynolds number. The Strouhal number
is an indication of the relative importance of local inertia
as compared to convective inertia.
An experimental curve which can be used to determine
the frequency of the vortex shedding is given in Fig. 13-22.
From this curve, one could determine the shedding fre-
quency which is useful information in practical design.
13-5.2 Drag Due to Unsteady Motion:
The Added Mass Concept
13-5.2.1 The concept of added mass, or virtual mass, or
induced mass, is of particular importance in the study of the
forces acting on a body accelerating or decelerating in still
water, or on a fixed body subjected to an unsteady current.
It is recalled that under steady state conditions the total
force acting on a fixed body by a current without circulation
is nil in the case of a perfect fluid. It is the paradox of
D' Alembert. In the case of a real fluid, it has been seen that
the force is a complex function of the Reynolds number.
Under unsteady conditions, another force has to be
added, whether the fluid be perfect or real. The value of
this force is analyzed below.
13-5.2.2 When a body of mass M moves in still water at a
speed U, it has a kinetic energy !MU 2 This body auto-
matically induces a fluid motion around it which tends
to zero when the distance from the body tends to infinity.
The exact law of decay depends upon the shape of the body.
However, far from the body it can be said that the fluid
particle velocity V(x,y,z,t) decreases as 1/R 3 in the case
of a three-dimensional flow and 1/R 2 in the case of a two-
dimensional flow; R being the distance of the considered
fluid particle from the center of the body.
The total kinetic energy of the fluid surrounding the
body is then
IIi:!pV (x,y,z,t) da
2
where lim is the limit of the body and da is an elementary
volume (or area in the case of a two-dimensional motion).
The total kinetic energy of the system can then be written
W= !uz[M + pIIi: (~Y da]
The quantity
M' = pIIi: (~Y da
is the added mass. It is the mass of fluid which, moving
at speed U, will have the same kinetic energy as the total
mass of fluid. W is the work that is required to give the
body speed U, or it is the work that would be required to
 Chapter 13: The Boundary Layer, Flow in Pipes, Drag, and Added Mass

100
175

10 ~~
Figure 13-20
Drag coefficient for circular cylinder
r-----1----
\_
0.1
-I I 2 3 4
10 10 10 10 10
R = VD
e II

stop it. It is seen that this work also includes the work also be in general a function of time. However, in the case
required to move the fluid around it: J:M'U 2 Once this of a perfect fluid, V(x,y,z,t)/U is independent of U, but
work is produced, the body will continue to travel in a depends upon the flow pattern only. If one refers V to a
perfect fluid at a constant velocity U. coordinate system moving at velocity U this ratio is also
independent of time. Hence, the integral of the coefficient
13-5.2.3 It is pointed out that because V decreases with V(x,y,z,t)/U is independent of the value of U and the time
R- 3 (or R- 2 in the two-dimensional case), (V/U) 2 varies as well. In a word, M' is a constant associated with the body
with distance as R- 6 (or R- 4 ), whereas the integral of dCJ and the specific mass of the fluid only.
varies as R 3 (or R 2 ). Consequently, the integral for M'
has a finite value. 13-5.2.4 Let us now consider the total force to move
It is seen also that in the general case M' is a function the body. It is equal to the sum of the inertia of this body
of the absolute value of U and consequently of the Reynolds
number U Djv and other empirical parameters charac-
Figure 13-22 Relationship between the Strouhal
terizing the flow (such as UT/D for periodic motion where D number and the Reynolds number.
is a typical dimension of the body). Consequently, M' will
a:: 022
w
ID
::< 0.20 /
Figure 13-21 Von Karman vortex street. ::>
z
...J 0.18
<( I
/
5 0.16

h
0
a::
0.14 I
I
~
CJ)

T CJ) 0.12
I 4
10 102 103 10
R~ REYNOLDS NUMBER
 Part 2: Some Mathematical Treatments of the Basic Equations

q6 itself and the inertia of the fluid surrounding it, i.e., should be determined from the force exerted by the fluid
on the body or vice versa, i.e.
JJ'foo dV
fi
dU
F = M dt + P Jlim dt da
F' = p cos eds
which can still be written as
e
where p is the pressure around the body, cos is the angle
F = (M + M')dU of the perpendicular to ds with the main direction of the
dt motion, and s the area of the body. (Fig. 13-23). Given
V (or), p can be determined by application ofthe Bernoulli
where
equation. In general, the integral of p V2 /2 being zero
M' = p(djdt) JJJ!im
V da (paradox of D'Alembert), the integral of p oc/Jjot only is
dU/dt significant, so that finally, replacing p by - pocjJjot

It is not evident a priori that the two definitions for M' M' = - Hp(oc/Jfot) cos e ds
are identical. As a matter of fact the integral dUjdt
Of course, the equality force-momentum can also be
obtained by differentiating the equality work-energy as
follows
may diverge as the distance from the body tends to infinity. !!__ (W) = !!__ [-!-(M + M')U 2 ]
Hence, in the case of moving body the force F' = M' dUjdt dt dt 2

POSITIVE
PRESSURE

Figure 13-23
An example of distribution ofpressure force
a)VELOCI T Y on an accelerating body in still water.
PRESSURE

c l TOTAL PRESSURE

b) ACCELERATION PRESSURE
 Chapter 13: The Boundary Layer, Flow in Pipes, Drag, and Added Mass

This gives This gives 177

dL , dU R2
F-= U(M+M)- V(r,e,t) = - 2 U(t)
dt dt r

Since dL/dt = U, the equality force-momentum is obtained. where U(t) is the velocity of the body. The total kinetic
Still, this operation is done by assuming M' constant and energy of the fluid per unit length of the cylinder is then
dFjdt = 0. Actually, a more rigorous demonstration will successively
not require this limitation.
In conclusion, for all practical purposes, the added
mass is determined by calculating the integral
T = f0
2" foo
r=R
1 R4U2
p -2 -
r
4- r dr de

M' = p JJJ(~)2 da
or in the case of an irrotational flow = tpnR2U2

, Hp(a<Pfat) cos e ds It is seen that the added mass isM' = pnR 2, i.e., the mass
M =- dU/dt of a cylinder of radius R having the same density as the
fluid. It is also seen that the total force to move the body is
Then this value of M' will be used for determining the force
M'(dU jdt). The shortcomings of these simple demon- F = (M + M') dd~
strations are not discussed in this book. It is only pointed
out that one of the limits of the integral, namely the body I.e.,
limit, could be time dependent.
dV
F = (pb + p)nRz dt
13-5.2.5 The case of a moving circular cylinder of radius
R is given here as an example. The velocity potential for a where Ph is the density of the body. It can be verified that
cylinder moving through a fluid at rest is given by super- F' = M'(dV/dt) is the total force exerted by the fluid on the
imposing upon the steady state pattern of a flow around a body as the sum of all the pressure forces in the direction
cylinder, a uniform velocity U (see Section 11-4.4.3) i.e., of the motion:

cp = - u(r + ~2 ) cos e + Ur cos e F' = f" P cos e R de

It is seen that this operation nullifies the uniform flow

The pressure distribution around a moving cylinder
component and the potential function is that of a doublet:
in the case of an unsteady motion is given by
R2
= -u-r cos e P
- - = -
a 1 2
+ - v + f(t)
P at 2
The fluid velocity at any point has a magnitude given by i.e.,
 Part 2: Some Mathematical Treatments of the Basic Equations

q8 Since the integral of the quadratic term is zero, the total It is seen that the pressure component due to local inertia
force acting on the cylinder is is in this case twice the value of the pressure component
in the case of a moving cylinder.
F' = 5 pR
0
2" dU
2 -
dt
cos 2 8 d8 =
dU
pnR 2 -
dt
Inserting this value in the previous integral yields

I.e., F = 2 nR 2 dV = 2M' dV
p dT dT
M' = pnR 2
13-5.2.8 In the case of a real fluid, this inertial force still
13-5.2.6 Let us now consider the case of a fixed body exists, but because of viscosity, separation, and wake,
subjected to an unsteady current. The total force exerted there is also a superimposed quadratic force. The following
by water on the body is still empirical formula is often proposed

F = fip cos eds F =

U2
pCvA-
2
+ pCMvol-dt
dU

which, in the case of an irrotational motion without circula-

where A is the cross section of the body perpendicular to the
tion of velocity is identical to
flow (A = 2R in the case of a cylinder), "vol" is the volume
of the body, Cv is the drag coefficient, and CM the inertial
coefficient. It is seen by comparison with the previous
result that C M = 2 in the case of a cylinder. As a matter of
This integral is twice the value of the same integral in the fact, both C M and C v are not constants but complex
case of a moving body in a still fluid, and consequently functions of the reduced frequency D2 jvT (Valensi number),
the Reynolds number UD/v, and time; (D 2 /vT)/(UDjv) is a
F = 2M'dV Strouhal number, which is also called the Iversen modulus.
dT The added mass of a floating body is also a function of
It is interesting to note that in the first case of a moving frequency, even in the case of a perfect fluid, as the waves
body in still water, the same force is found provided generated by the body is a function of frequency.
M = M', i.e., the body has the same average density as the
fluid. This is true of floating systems and submarines. 13-5.2.9 The previous formula is used by engineers to
calculate the wave force on small vertical piles. The hori-
13-5.2.7 The case of a fixed circular cylinder subjected zontal velocity component u(x,z,t) then replaces U, and
to an unsteady fluid flow is given here as an example. the formula is called the Morison formula. Assuming that
The potential function for the motion is then the pile disturbance on the velocity field is small, u is ob-
tained by a wave theory. It is not entirely clear if the experi-
cf> = - U(t)(r + ~2 ) cos e ment in unidirectional acceleration of a cylinder through a
fluid is strictly applicable to the wave force problem. Also the
mass transport due to wind stress near the free surface
and
increases considerably the particle velocity as compared
-I
to the particle velocities given by classical wave theories
p81> dU
=-p2R-cose
8t r=R dt (see Section 16-1.5).
 Chapter 13: The Boundary Layer, Flow in Pipes, Drag, and Added Mass

13-5.2.10 In the case in which both a body moves with
a time dependent velocity V(t), in a time dependent fluid
flow, which has a velocity U(t) at a distance from the body,
the total force to be exerted to move a body of mass M is
F = M dV + M' !!__ (V - U)
dt dt
13-5.3 The Case of Time-Dependent
Mass
13-5.3.1 In the case of (for example) an underwater
missile popping up vertically from the tube of a submarine,
and subjected to a steady current U, an inertial force
acting in the direction of the current is added to the drag
force. This added force is due to the variation of mass of
the body subjected to the cross-flow, i.e., the momentum
equation is now written
F = dmU = Udm
dt dt
where dmjdt is proportional to the rate of volume of missile
exposed to the fluid flow, i.e., the product
The first term results from the added mass, as defined in 179
Section 13-5.2 (vide supra). The second term is the result of
the free surface variation and is equal to
nD 2
pCMus4Ws
where CM is a dimensionless coefficient, us and ws are the
free surface horizontal and vertical particle velocity
components at the pile location (ws = aiJ;at, l](t) being
the free surface elevation), and D is the pile diameter.
This force is often neglected. Actually, this neglect intro-
duces a large error on the calculation of wave force and
torque, in the case of piles subjected to large-amplitude
shallow water waves.
PROBLEMS
13.1 The dissipation energy thickness ()** of a boundary
layer is defined by the equation:

J AdV~
nD 2 pU6 (5** = p Ln u(U6 - u2 ) dy
4 v

where the right-hand term is the flux of dissipated energy

where Vis the exit velocity of the missile and A is the cross by friction. Calculate the value of J*, 8, and()** as functions
section of the missile of diameter D. of (5 in the cases where
Therefore,
1. u = U 0 ~ for y < (5 and u = U 0 for y :;::: (5

where CM is a dimensionless coefficient. 2. u = U{ 2 ~- GYJ for y < (5 and

13-5.3.2 In the case of fixed vertical piles subjected to u = U 0 for y :;::: c5

wave action, the fluid velocity is time dependent and,
owing to the variation of elevation of the free surface, the
volume of pile into the water also time dependent. In this
case:
13.2 Obtain the transverse velocity component v for the
laminar boundary layer along a plate.
13.3 Determine the coefficients An of the Blasius theory
up to A 11 as a function of A 2 . Demonstrate that only
 Part 2: Some Mathematical Treatments of the Basic Equations
180 A 3 n+ 2 are different from zero and establish a reference
formula for A 3 n+ 2 as a function of A 2 . Present the expression
of f(ry) as a power series as a function of A 2 and determine
the value of A 2 (it is found that A 2 = 0.332). Determine the
value of f(ry), f'(ry), f"(ry) at y = 0.
13.4 The thickness of the laminar boundary layer on a
semiflat plate can be evaluated through the von Karman
momentum integral formula by assuming a proper velocity
profile. If the velocity profile is assumed to be a polynomial
1. Give the proper boundary conditions and use these
boundary conditions to determine the five constants.
2. Calculate the shear stress along the plate.
3. Obtain the thickness of boundary layer by use of
von Karman's momentum integral formula.
13.5 Consider the steady boundary layer on a convergent
channel with flat walls (Fig. 13-24). The boundary-layer
equations along the wall parallel to the x axis are
au au 1 op o2 u
u-+v-=---+v~
OX oy pOX oy 2
op = 0 au + ov = 0
ay ax ay
The free-stream velocity is given in the form u(x) = -(u 0 /x).
Introducing the similarity transformation
as well as the stream function
1. Find the ordinary differential equation for the stream
function and the boundary conditions.
2. Obtain the velocity distribution.
Figure 13-24
13.6 Establish the momentum integral equation for
unsteady boundary layer.
13.7 Determine the frictional force on an oscillating plane
covered by a layer of fluid of thickness h. The frequency
of oscillation is k and the fluid has kinematic viscosity v.
13.8 A fixed amount of discharge Q flows uniformly down
a semiinfinite plate started at x = 0. The fluid has viscosity
J1 and density p and it accelerates with the gravitational
acceleration g as shown in Fig. 13-25. The free surface is
assumed to be of constant pressure. The fluid in contact
with the plate forms a boundary layer. The thickness
of the boundary layer increases until it reaches the free
surface as shown in Fig. 13-25. The flow will continue
to accelerate and the fluid layer will become thinner until
it reaches an asymptotic value.
1. Derive the momentum integral equation before the
boundary layer reaches the free surface.
2. Assume a parabolic velocity profile
u: = ao +at(~)+ a2(~Y
Determine the constants a0 , at,
a 2 by use of appro-
priate boundary conditions. Substitute the determined
velocity profile into the momentum integral equation
and derive a differential equation for b(x).
3. Try a solution of the form c5 = f3xn. Determine the
values of f3 and n from the integral equation.
 Chapter 13: The Boundary Layer, Flow in Pipes, Drag, and Added Mass

PO 13.9 Derive the resistance equation for turbulent flow in 181

rough pipes
\ 1 r0
\ u.=~ ! 112 = 2log 10 ks + 1.74
\
\ 13.10 The potential function for a two-dimensional flow
\ around a cylinder of radius R is

= u(r + ~2 ) cos e
p =p = CONSTANT
0 where U is the velocity at infinity. Give the pressure distri-
bution around the cylinder in the case where
1. U = constant = U 0
2. U = U 0 sin kt
Determine the total force acting on a cylinder by integration
of the pressure.

13.11 Calculate the added mass for a sphere, taking into

account the fact that the velocity potential for a sphere
of radius R moving at velocity U in a spherical coordinate
system (r,8,<1>) in a still fluid is
VR 3
X
= 2r2 cos e
Figure 13-25

13.12 The horizontal velocity component due to a linear

4. Determine the distance x 0 at which the boundary layer
reaches the free stream.
5. Determine the thickness of the boundary layer !5 0 at
the location where the boundary layer reaches the free
stream.
6. Derive the momentum integral equation for the flow
regime x > x 0
7. Using the velocity profile derived in question (2),
derive the differential equation of J(x) for x > x 0 .
8. Obtain the relationship between the distance x and
the layer thickness.
9. Obtain the layer thickness !5 3 when x approaches
infinity.
periodic gravity wave in deep water is
Hn
u = - e 2 "z/L cos (kt - mx)
T
Calculate the maximum total force exerted on a vertical
cylinder of 5 ft (152 em) diameter by a wave H = 20 ft
(609 em), T = 10 sec. The drag coefficient Cn = 1, and the
inertial coefficient CM = 2, k = 2n/T, m = 2n/L and
L = gT 2 /2n.
Chapter 14 14-1 Uniform Flow, Normal Depth,
and Critical Depth
14.1.1 Hydraulic Radius and the Chezy

Open-Channel Formula
14-1.1.1 Three kinds of approaches are used to obtain

Hydraulics the equation used in open-channel hydraulics. However,

all of these mathematical forms originate from the Newton-
ian equation, and a criterion for choosing one rather than
another does not really exist.
Most of the basic equations of this chapter can be ob-
tained from (1) a simplified form of the Eulerian equations
in which an empirical friction term is included, (2) the
generalized Bernoulli equation in which the condition at the
free surface p = Pa (atmospheric pressure) is inserted, or (3)
by a direct application of the balance of forces in which the
significant terms only are included.

14-1.1.2 Consider a uniform flow parallel to the axis OX

(see Fig. 14-1). Since the motion is steady, o(u,v,w)!at = 0,
and since the motion is uniform, v and w = 0, and oujox = 0.
Hence, it is easily verified that all the inertial terms are zero.
Moreover, the pressure forces acting on each side of cross

Figure 14-1 Uniformjiow.

 Chapter 14: Open-Channel Hydraulics

section A of an element of fluid A .1x are in equilibrium. equation for Vis obtained r83
The OZ components of the pressure force and gravity also
balance independently of the flow velocity. Thus, the only
significant forces which remain are the gravity component
V = (y) 112
(RH sin 8) 112

in the 0 X direction and the shearing forces: The Chezy coefficient of dimension (LT- 2 ) 1 12 is Ch =

J:
(g/!) 1 12 ; thus f = g/C~. The slope S is generally small, so
pgA .1x sin (} = .1x < dP that S = tan(} ~ sin(} and V = Ch(RHS) 112 This is the
Chezy formula.
where P is the "wetted perimeter" (i.e., the length of the The discharge Qn = VA is then
perimeter of the cross section A which is underwater) and r Qn = ACh(RHS) 112 =KS 112
is the shearing stress per unit area.
where K = ACh(RH) 112 is the conveyance of the channel
and depends upon the geometry of the cross section of the
14-1.1.3 In general, r is not constant, except around a
channel and the water depth only. Qn is the normal discharge,
circular pipe due to symmetry. However, owing to secon-
which is defined as a function of water depth for a given
dary currents, the variation of r can often be considered as
channel.
negligible (see Section 8-2.3).
The previous equation is then written 14-1.15 The Chezy coefficient and f can only be deter-
mined by experiment. It is found that in the case of turbulent
A
pgRH sin (} = r where RH=- flow over a rough bottom
p
- 1.486 1/6
Ch ---RH
Note that the "hydraulic radius" RH has the dimension of a n
length. It is easily verified that in the case of a rectangular where R H is in feet and n is the Manning coefficient; n is
section of depth h and width l given as a function of relative roughness and in practice
lh varies between 0.01 and 0.05. Inserting this expression in the
RH = l + 2h Chezy formula gives the Manning formula

and RH tends to h when l --+ oo; i.e., in the case of a large V = 1.486 Ri/3S1/2
n
river. In the case of a circular pipe of radius R
and the conveyance
nR 2 R
RH=--=- K = 1.486 ARi/3
2nR 2 n

14-1.1.4 In the case of a river or a channel, the Reynolds
number is generally large, and the flow is fully turbulent.
The shearing stress can then be assumed to be related to the
average velocity V by a quadratic function such as
r = pfV2
where f is a dimensionless friction factor. When this is
combined with the equation pgRH sin(} = r the following
14-1.2 Normal Depth and Transitional
Depth
14-1.2.1 The normal depth hn is defined as the distance
between the lowest part of the channel and the free surface
of a uniform flow. It is determined by the equality
$~2 = K(hn)
Part 2: Some Mathematical Treatments of the Basic Equations

where K(hn) is the function that characterizes the conveyance this formula because Vis average velocity. For the sake of
of the channel. In the case of a wide rectangular channel, simplicity, oc will be neglected in the following discussion.
since RH = hn, the normal depth hn is It is seen that the specific energy is the sum of the kinetic
qn
( chsl/2
)2/3 energy (V 2 /2g) and the potential energy per unit of weight
with respect to the bottom of the channel h (but not with
hn =
respect to a horizontal datum) (see Fig. 14-2).
where qn is the normal discharge per unit width. It is seen The specific force is

ff (-v2g + -pgP) dA = -gA

that, in the case of a uniform flow and for a given channel,
Q2
there is a unique relationship between normal depth and I = + aA
A
normal discharge.
where a is the distance between the center of gravity of the
14-1.2.2 In the case of a nonuniform flow, the term cross section A and the free surface. It is recalled that the
djdx(V 2 /2) is no longer zero; hence gravity force and friction specific force is the sum of the momentum per unit time and
force do not balance exactly. The water depth h is different per unit weight and the integral of pressure force to a cross
from the normal depth; h is then a "transitional depth," section of the channel per unit weight.
which varies with distance and can be larger or smaller In the case of a two-dimensional channel (rectangular
than hn. For example, the transitional depth upstream of a cross section), A = hl, a = h/2, and q = Q/l; thus E and I
dam is larger than the normal depth. per unit width become
It is assumed that the values or formulas for the friction q2
coefficients J, Ch, and n obtained in the case of a steady E = 2gh2 + h
uniform flow with normal depth are still valid in the case of
a nonuniform flow with a transitional water depth. In q2 h2
I=-+-
general, the lack of accurate information on the value of gh 2
the friction coefficient makes this approximation compat- respectively. Writing
ible with the inherent error in the original formula. Con-
sequently, the conveyance K is a general function of h h* = !!_ *- q
q - E(2gE)lf2
which can also be used for nonuniform flow. E'
one obtains the equation
14-2 Specific Energy, Specific Force, q* = h*(1 - h*) 112
and Critical Depth which is universal.

14-2.1 Definition of the Specific Energy Figure 14-2 Notation for an open channel.
and Specific Force
The quantity
_ (1 + oc) V 2 h _ (1 + oc)Q 2 h
E- 2g + - 2gA 2 +
where V is the average velocity, h the maximum water
depth, Q the discharge, and A the cross section, is called the
specific energy; oc is a positive coefficient which appears in
 Chapter 14: Open-Channel Hydraulics

h h

I q INCREASES
I
/,

Ec E = :-i!_ +h !=
v2
9fl+2
h2
2g
Figure 14-3 Specific energy and specific force.

14-2.2 Definition of the Critical Depth 14-2.2.3 In the case of a complex cross section A(h) the
critical depth is defined by
14-2.2.1 Let us consider the two functions E(q,h) and
l(q,h) in the case of a rectangular channel. It is seen that for aE 2 Q2 aA
a given value q, E(h), and J(h) vary as shown on Fig. 14-3a,b. oh = - A 3 29 oh = 0
1
In particular, the minimum values for E and I given by Since oAjoh = l, l(h) being the width of the river at the
oEjoh = 0 and oljoh = 0 are obtained for the same value free surface, one has
of h = h0 when Qz A3
g
Because A and l are functions of h only, it is then possible to
establish the curve he as a function of the critical discharge
i.e., Qe for any kind of cross section (see Fig. 14-4). It is easily
verified that in the case of a rectangular cross section, the
he = ( qg2)1/3 or V = (ghe)l/2 previous value he is found.

in which case Figure 14-4 Variation of the critical depth with discharge.

14-2.2.2 Similarly, the function q(h) for a given value

E = constant varies as shown in Fig. 14-3c. It is easily
verified that the maximum value for q given by oqjoh = 0,
is also found to be for h = he.
 Part 2: Some Mathematical Treatments of the Basic Equations

r86 i.e.,
14-3 Tranquil Flow and Rapid Flow g A

14-3.1 Definition
se = c; RHt
and in the case of a large rectangular channel Sc = gjC~ = f
There exist two possible values, h 1 and h 2 , for the same
When S > So one has hn < he corresponding to a value of
values of E and q, defined by
the Froude number V/(gh) 1 12 < 1: the flow is subcritical or
Vi v~ tranquil. When S < So one has hn > he corresponding to a
E = h1 + - = h2 + - value of the Froude number Vj(gh) 112 > 1, the flow is
2g 2g
supercritical or rapid.
and
14-3.3 Free Surface Disturbance and
Flow Control
such that Since any disturbance travels at an approximate speed
(gh) 112 , the occurrence of V > (gh) 112 means no perturba-
tion can travel in an upstream direction. Also, when the
flow characteristics change from a tranquil upstream flow
to a rapid downstream flow such as at the top section of a
By inserting the value for he = (q 2 jg) 113 , one obtains
weir, the discharge is controlled at this critical cross section.
2hih~ = h3
h2 + hl e
14-4 Gradually Varied Flow
The larger depth, say h 1, corresponds to a "subcritical" or
"tranquil flow" in which case V1 < (gh) 112 The smaller 14-4.1 Basic Equations
one, say h2 , corresponds to a "supercritical" or "rapid 14-4.1.1 The generalized Bernoulli equation with friction
flow" in which case V2 > (gh) 112 can be applied to a gently sloped free surface flow as
follows:
14-3.2 Critical Slope
It has been seen that the normal discharge is a function
~ (v2 + !!._ + z) =
dx 2g pg
of the conveyance and the bottom slope: Qn = K(S) 112 ,
whereas the critical discharge is a function of the cross I.e.,
section and the width of the channel only Qe = (gA 3 j/) 112
There is a slope, the" critical slope," for which the normal
discharge, i.e., the discharge of a uniform flow, is equal to
the critical discharge, and consequently the normal depth where
hn is equal to the critical depth he. Q2 Q2
Consequently, the critical slope is defined by --=c"'~R-=-H-A"z = -K-2 = Sf
S1 is the slope of the energy line or head loss. Then, insert
p = Pa (atmospheric pressure) as a constant and call z the
 Chapter 14: Open-Channel Hydraulics

elevation of the free surface with respect to a horizontal one has finally,
datum. Since dh 1 - (Q 2 jK 2 S)
dz dh dx = S 1 - (Q 2 ljgA 3 )
-S+-
dx dx which is the fundamental equation for gradually varied
(see Fig. 14-5) the result is flow. Considering that

and

this equation can be written

The quantity in parentheses is recognized to be the specific dh = S 1 - (SI/S)
energy.
dx 1 - (SJ/Sc)
In the case of a large rectangular channel (RH ==o h) one
14-4.1.2 Since
obtains, after some elementary operations,
dh h3 - h3
dx = S h;-
c
h3

Figure 14-5
Notation for gradually varied flow .

DATUM

: ''':"!2~0.',i;c(T'''~;;:y~.;~'<;\l'h"'\,~;~i'ifu~t%;'i<;~;;c),~:r-<\(9
tJ. X
 Part 2: Some Mathematical Treatments of the Basic Equations

188 Table 14-1 Backwater curves

14-4.2 Backwater Curves Areas Water
14-4.2.1 In general, the free-surface variation depends depth
upon (1) the sign of S; (2) the sign of the numerator, depend- Bottom slopes A B c Water depths variations
ing upon the water depth h by comparison with the normal Horizontal h > hn >he
depth hn; and (3) the sign of the denominator, depending (S = 0) Hz hn > h >he D
upon the water depth h by comparison with the critical B
H3 hn >he> h
depth he.
It is to be noted that h > hn or h < hn may involve h > he Adverse h > lhnl >he
or h < h<> depending whether hn > he or hn < he. In (S < 0) Az lhnl > h >he D
practice, the first calculation will consist of comparing the A3 lhnl >he> h B
value of hn and he.
For example, consider the case of a flow in a channel Gentle M, h > hn >he B
with a positive slope: S > 0, in which one finds hn > he.
(0 < S < SJ Mz hn > h >he D
In the case in which h > h", both the numerator and denom- M3 hn >he> h B
inator are negative, so dhjdx is positive. The water depth Critical c, h >he= hn B
continuously increases with distance. It is the case of a (S = SJ u
Cz he= h = hn
backwater curve due to a dam on a river with a gentle
c3 he= hn > h B
slope (see Fig. 14-6, case M 1).
Table 14-1 and Fig. 14-6 illustrate all the cases that may Steep s, h >he> hn B
be encountered. (S > Sc > 0) Sz he> h > hn D
Area A corresponds to the case in which the transitional s3 he> hn > h B
water depth is larger than both the normal depth h" and the
critical depth he. Area B corresponds to the case where the Notation as shown in Fig. 14-6.
transitional water depth is between hn and he. Area C
corresponds to the case where the transitional water depth
of a finite difference method over a succession of intervals
is smaller than both hn and he. In depth variation B (back-
,1x. Also, a number of approximate methods and graphical
water) means that the water depth tends to increase in the
methods exist. A number of tables and graphs can be
direction of the flow. D (draw-down) means that the water
found in the technical literature for a dimensionless
depth tends to decrease in the direction of the flow. U
"channel unity." As usual, the most direct approach based
(uniform) means that the water depth is constant.
on a finite difference method is the most practical and can
All other cases can be found systematically by a simple
be easily programmed for high-speed computers.
analysis of the basic equation. Similarly, the backwater
curves on channels of variable slope, continuous or dis-
continuous, can be analyzed and calculated. 14-5 Rapidly Varied Flow

14-4.2.2 In engineering practice, the calculation of 14-5.1 Hydraulic Jump

backwater curves can be done either by an exact integration 14-5.1.1 When the flow is rapid upstream and tranquil
in some simple cases such as a rectangular channel (the downstream, the transition is done through a hydraulic
Bresse formula), or most often numerically by application jump. Let us consider the simple case of a rectangular
 Chapter 14: Open-Channel Hydraulics

AREA A AREA B: AREA C:

h > hn ; h > he
w ... ...
I I
<1.
0
_j
en hn NONE

I'
_j

~
0z u he
N .c;
ii:A ' '
0 c
J: .c;

w
Figure 14-6
Different kinds of backwater curves.
<1.
0
_j
.-MI
l -
~-
M2
-
en
hr r
w u
n hn
J
_j.c;
he he .......r~c hn
li;:A .L M3
w c
(!).C:
' ' I
I .. L
~- '

w
<1.
0
_j
en
_j
<( u
u.c:
i= II
fj ~c

w
<1.
0
_j
en
u
<1. .c;
w
wv
1- c
en .c:

channel of width unity. The discharge is q. The momentum
theorem applied to the hydraulic jump is then (see Section
12-3.3):
where h 1 and h2 are the "conjugate" depths. For every
upstream value h 1 < h"' there corresponds a value h2 > he.
The backwater curve h'1(x) for rapid flow is calculated by
application of the Bernoulli equation from the upstream
conditions. The backwater curve for tranquil flow h2(x)
is also determined by application of the Bernoulli equation
Part 2: Some Mathematical Treatments of the Basic Equations

--I
CONJUGATE
OEPTH
hI
--
.........

\ TRANQUIL FLOW
Figure 14-7
/ Determination of the location of

--
an hydraulic jump.

HAP/0 FLOW .. ~

from the downstream conditions. The hydraulic jump one obtains

takes place between these two curves at a location where 11H = (h2 - ht)3
the two water depths are conjugated. Therefore, it is suffic- 4hlh2
ient to calculate for each value h 1 < he the corresponding
conjugate depth h~ > he (Fig. 14-7). The hydraulic jump and the energy loss per unit time is
takes place at the location where the backwater curve h2 (x) dE
for tranquil flow, crosses the curve h'1(x), conjugate of dt = pgq 11H
h 1(x).
The location of such a hydraulic jump is somehow 14-5.2 Effect of Flow Curvature
unstable. It could be stabilized by an obstacle, or a sudden The pressure distribution is no longer hydrostatic
deepening. Such a system is used in water power installa- in the case of rapidly varied flow. The path curvature has a
tions as a stilling basin for dissipating energy downstream nonnegligible influence on the flow behavior. Furthermore,
spillways or bottom outlets. a uniform flow near critical depth is very unstable. The
flow motion may then enter the category of flow motion
14-5.1.2 The head loss caused by a hydraulic jump is with path curvature.
determined by the difference Call R the radius of curvature of the free surface, such
that
1 d 2hjdx 2 d 2h
R = [1 + (dhjdx) ] = dx 2
2 312
Assume that the curvature is linearly distributed from
which gives the bottom (z = -h) to the free surface, such that
1 h + z d2 h
---
R(z) h dx 2
The centrifugal acceleration is V 2 / R where V is the average
Inserting the hydraulic jump equation (see Section 12-3.3)
velocity and the dynamic equation along a vertical becomes

~~ _ pg + Pv 2 [h : z ~:~ J

which, when integrated, is
p(z) = -pgz v2 d2h [ hz +-
+ p~--
h dx
2
z
2
2
J discharge, and Qc the critical discharge. Also demonstrate
Boussinesq has made an extensive use of this equation for
determining the condition when a flow motion is stable or
may become undulated, or when it results in an undulated
hydraulic jump.
Also, the study of flow over a rapidly variable slope such
as over a weir has been investigated by taking flow curva-
ture into account. This subject will not be developed further
in this book.
PROBLEMS
14.1 Determine the value of the hydraulic radius Ra for a
circular conduit as a function of the maximum water depth.
14.2 A trapezoidal channel has a width at the bottom of 40ft
(12.19 m) and a bank slope of 45. The bottom slopeS =
0.002, and the Manning coefficient n = 0.03. Calculate the
curve for a normal depth y n as a function of the discharge Q"
up to Q = 1000 ft 3 jsec (28.32 m 3 jsec). Calculate the curve
giving the critical discharge Qc as a function of the critical
depth Yc and determine the value of the discharge for which
critical depth is equal to normal depth.
14.3 It is possible to define a channel that has a constant
hydraulic radius Ra whatever the water depth h?
14.4 Demonstrate that the free surface slope of a steady
gradually varied flow in an open channel is equal to the
sum of the energy slope and the slope due to the velocity
change d(V 2 j2g)jdx.
14.5 Demonstrate that
dh = S 1 - (Q/Qn) 2
dx 1 - (Q/Qc) 2
Chapter 14: Open-Channel Hydraulics
where S is the bottom slope, Q the discharge, Q" the normal
that, in the case where the Manning formula is used for
large rectangular channels,
dh 1 - (hn/h) 1013
dx = S 1 - (hc/h) 3
Demonstrate also that in a rectangular channel of variable
width /(x)
dh S - Sf + (Q 2 /gA 3)dljdx
dx 1 - (Q 2 l/gA 3 )
14.6 From the equation
dh h3 - h3
dx = S h3 - h~c
and depending upon the sign of S and the relative value of
h, hn, and h0 determine all the possible kinds of backwater
curves that may be encountered in a rectangular channel.
14.7 Determine all the kinds of water surface profiles that
are encountered in the two cases in which the normal depth
is larger than the critical depth and at the opposite, where
the critical depth is larger than the normal depth in Fig.
14-8, next page. The opening of the gate 0 will be considered
as a variable in such a way that 0 can be smaller or larger or
both than he and hn.
14.8 Establish under which conditions there are two normal
depths and two critical depths in a circular closed conduit.
Determine all the possible backwater curves that may exist
in such a conduit.
14.9 Establish the required condition for establishing a
hydraulic jump on a sudden bottom drop corresponding
to various assumptions on the exact location of the jump
with respect to the sudden bottom drop.
14.10 Demonstrate that the ratio of gain of potential energy
defined by the difference h 2 - h 1 to the loss of kinetic
Part 2: Some Mathematical Treatments of the Basic Equations

sz= UPSTREAM LEVEL GATE

Figure 14-8

DOWNS TREAM
sz= LEVEL

energy defined by the difference REFERENCES FOR PART TWO

Vi- v~ Bakhmeteff, B. A., Hydraulics of Open Channels. McGraw-Hill,
2g New York, 1932.
across a hydraulic jump in a rectangular channel is Caratheodory, C., Conformal Representation. Cambridge University
Press, London, 1941.
4hlh2 Chow, V. T., Open Channel Hydraulics. McGraw-Hill, New York,
(hl + h2) 2 1959.
Hinze, J. 0., Turbulence. McGraw-Hill, New York, 1959.
14.11 Demonstrate that the efficiency of a hydraulic jump, Kellogg, 0 . D., Foundations of Potential Theory. Murrey Printing
defined as the ratio of the specific energy after and before Co., New York, 1929.
the jump, is Kober, H., Dictionary of Conformal Representations. Dover Pub-
lications, Inc., New York, 1952.
E2 (8Fi + 1) 312 - 4Fi +1 Lamb, H., Hydrodynamics, 6th ed., Dover Publications, New
E1 8Fi{2 + Fi) York, 1945.
where Landau, L. D., and Lifshitz, L. D., Fluid Mechanics. Pergamon
Press, London, 1959.
vl Lin, C. C., The Theory of Hydrodynamics Stability. Cambridge
F = (ghl)l /2 University Press, London, 1955.
McNown, J. S., Hsu, E. Y., and Yih, C. S., Applications of the
Subscripts 1 and 2 refer to upstream and downstream relaxation technique in fluid mechanics. Trans. , ASCE, 120:
values, respectively. 650- 686, 1955.
Milne-Thomson, L. M., Theoretical Hydrodynamics, 2nd ed.,
14.12 Consider a vertical jet hitting a horizontal large Macmillan, New York, 1950.
circular plate at its center. The plate is ended by a weir Prandtl, L., Essentials of Fluid Mechanics, Hafner, New York, 1952.
providing a quasiconstant water depth h. Determine the Rouse, H., editor, Advanced Mechanics of Fluids. Wiley & Sons,
location of the circular hydraulic jump as a function of the New York, 1959.
discharge Q of the jet and h. The bottom friction forces will Rouse, H., editor, Engineering Hydraulics, Proceedings of the 4th
be neglected. Hydraulic Conference, Iowa, Wiley & Sons, New York, 1949.
 Chapter 14: Open-Channel Hydraulics

Sabersky, R. H., and Acosta, S. A., Fluid FlQw-A First Course in 193
Fluid Mechanics. Macmillan, New York, 1964.
Schlichting, H., Boundary Layer Theory. McGraw-Hill, New York,
1960.
Streeter, V. L., editor, Handbook of Fluid Mechanics, McGraw-Hill,
New York, 1961.
Taylor, G. I., Statistical theory of turbulence. Proc. Roy. Soc. Lond. A,
151, 1935.
Townsend, A. A., The Structure of Turbulent Shear Flow. Cambridge
University Press, London, 1956.
Von Karman, T., Collected Works. Butterworth Scientific Publica-
tions, London, 1956.
Von Karm{m, T., and Howarth, L., On the statistical theory of
isotropic turbulence. Proc. Roy. Soc. Lond., 164, 1937.
 PART THREE

Water Wave Theories

 Chapter 15
197
15-1 A Physical Classification of
Water Waves and Definitions

15-1.1 On the Complexity of Water

An Introduction to Waves
15-1.1.1 The aim of this chapter is to present the theories
Water Waves for unsteady free surface flow subjected to gravitational
forces. Such motions are called water waves, although
pressure waves (such as acoustic waves) in water are also
water waves. They are also called gravity waves, although
atmospheric motions are also waves subjected to gravity.
From the physical viewpoint, there exist a great variety
of water waves. Water wave motions range from storm
waves generated by wind in the oceans to flood waves in
rivers, from seiche or long period oscillations in harbor
basins to tidal bores or moving hydraulic jumps in estu-
aries, from waves generated by a moving ship in a channel
to tsunami waves generated by earthquakes or to waves
generated by underwater nuclear explosions.
From the mathematical viewpoint, it is evident that a
general solution does not exist. Even in the simpler cases,
approximations must be made. One of the important
aspects of water wave theories is the establishment of the
limits of validity ofthe various solutions due to the simplify-
ing assumptions. The mathematical approaches for the
study of wave motion are as varied as their physical aspects.
As a matter of fact, the mathematical treatments of the
water wave motions embrace all the resources of mathe-
matical physics dealing with linear and nonlinear problems
as well. The main difficulty in the study of water wave
motion is that one ofthe boundaries, namely the free surface,
is one of the unknowns.
Water wave motions are so varied and complex that any
attempt at classification may be misleading. Any definition
corresponds to idealized situations which never occur
rigorously but are only approximated. For example, a
pure two-dimensional motion never exists. It is a convenient
mathematical concept which is physically best approached
in a tank with parallel walls. Boundary layer effects and
Part 3: Water Wave Theories
transverse components still exist although they are difficult
to detect.
15-1.1.2 It has to be expected that, due to this inherent
complexity, a simple introduction to the problem of water
waves is a difficult, if not impossible, task. Hence, this
chapter should rather be regarded as a guide for the
following chapters and for continuing further study beyond
the scope of this book.
The full assimilation of the subject leading to a clear-cut
understanding of this chapter can only come after a compre-
hensive study of each existing theory within or beyond the
scope of the present book. With this in mind, the following
classification is proposed. A physical classification is given
first, then the different mathematical approaches and their
limits of validity are introduced. Finally, the traditional two
great families of water waves are presented.
15-1.2 Oscillatory Waves
From the physical viewpoint, there exist essentially two
kinds of water waves. They are the oscillatory waves and
the translatory waves. In an oscillatory wave, the average
transport of fluid; i.e., the discharge or rpass transport, is
nil. The wave motion is then analogous to the transverse
oscillation of a rope (see Fig. 15-1). A translatory wave
involves by definition a transport of fluid in the direction
in which the wave travels. For example, a moving hydraulic
jump, so called tidal bore or simply bore, is a translatory
wave.
15-1.2.1 An oscillatory wave can be progressive or
standing. Consider a disturbance I'J(x,t) such as a free
WAVE DIRECTION
surface elevation traveling along the OX axis at a velocity C.
The characteristics of a progressive wave remain identical
for an observer traveling at the same speed and in the same
direction as the wave (Fig. 15-2). In the case where 11 can
be expressed as a function of (x - Ct) instead of (x,t), a
"steady state" profile is obtained. I'J(X- Ct) is the general
expression for a steady progressive wave traveling in the
positive 0 X direction at a constant velocity C. In the case
where the progressive wave is moving in the opposite
direction, its mathematical form is expressed as a function
of (x + Ct). It is pointed out that the definition of wave
velocity C for a nonsteady profile is not strictly valid, since
each "wave element" travels at its own speed, resulting in
wave deformation.
15-1.2.2 The simplest case of a progressive wave is the
wave which is defined by a sine or cosine curve such as
H {sin}
1'/ =2 cos m(x - Ct)
Such a wave is called a harmonic wave where H/2 is the
amplitude and H the wave height.
The distance between the wave crests is the wavelength
L, and L = CT where T is the wave period. The wave-
number (or number of cycles per unit distance) m = 2n/L
is the number of wavelengths per cycle. The frequency is
k = 2n/T. Hence the previous equation can be written
1'/ = 2H{sin}
cos 2n
(x t)
L - T
H/L is the wave steepness.
Figure 15-1
Oscillatory wave.
 Chapter 15: An Introduction to Water Waves

199

Figure 15-2 o IC--....L.::::._--+------=:::~F=~~I-------::.L----+---t-

Progressive wave.

15-1.2.3 A standing or stationary wave is characterized A standing wave generated by an incident wind wave is
by the fact that it can be mathematically described by a called clapotis. In relatively shallow water (d/L < 0.05) it
product of two independent functions of time and distance, is called a seiche. A seiche is a standing oscillation of long
such as period encountered in lakes and harbor basins. The ampli-
tude at the node is zero, and at the antinode it is H.
,~,. H . 2nx . 2nt
'I'= sm-sm-
L T 15-1.2.4 Two waves of same period but different ampli-
or more generally: tudes traveling in opposite directions form a "partial
clapotis" and can be defined linearly by the sum of A sin
(x - Ct) + B sin (x + Ct). A partial clapotis can also be
A standing wave can be considered as the superposition considered as the superposition of a progressive wave
of two waves of the same amplitude and same period with a standing wave. A partial clapotis is encountered
traveling in opposite directions. In the case where the in front of an obstacle which causes partial reflection.
convective inertia terms are negligible, the standing wave
15-1.2.5 It will be seen that the wave velocity is in general
motion is obtained by a mere linear addition ofthe solutions a function of the water depth (see Section 16-3.2). Hence, if
for two progressive waves (see Fig. 15-3). The following the depth remains constant, wave crests remain parallel.
identity holds:
However, when the depth varies in a direction not parallel
H 2n H . 2n to the wave crest, different parts of the crest travel at
2 sin L (x - Ct) +2sm L (x + Ct) different speeds, and the direction of propagation of the
crest changes. This is the phenomenon of refraction which is
. 2n 2n encountered when a wave travels from one water depth to
= H sm L x cosT t
another water depth at an angle with the bottom contours.

Figure 15-3
Standing wave.
 Part 3: Water Wave Theories
200
WAVE HE IGHT
WAVE CREST
BREAKWATER
mmzzmz{{;,-,-------~,~mmzzzz
15-1.2.6 The phenomenon of diffraction is encountered
at the end of an obstacle (Fig. 15-4). It can be considered as
a process of transmission of energy in a direction parallel
to the wave crest although this definition oversimplifies a
more complex phenomenon. The phenomenon of wave
diffraction is actually the result of phase differences be-
tween the waves radiating from a line source, which is
interrupted by the obstacle. It is an end effect.
15-1.2.7 The breaking phenomenon is encountered at
sea under wind action (white caps), on beaches (surf), and in
tidal estuaries (tidal bores) (see Fig. 15-5).1t is a shock wave
phenomenon which is also encountered in gas dynamics.
The breaking phenomenon is characterized by a high rat$!
of free turbulence and air entrainment associated. with a
high rate of energy dissipation. Bores generated by wind
waves breaking on beaches or by tides of high amplitude
in estuaries should be regarded as translatory waves
rather than oscillatory waves.
A number of equivalent definitions can be given for the
breaking criterion: Breaking occurs (1) when the particle
velocity at the crest becomes larger than the wave velocity,
Figure 15-4
Wave diffraction.
(2) when the pressure at the free surface given by Bernoulli
equation is incompatible with the atmospheric pressure,
(3) when the particle acceleration at the crest tends to
separate the particles from the bulk of the water surface, or
(4) when the free surface becomes vertical. Accordingly, the
following theoretical formulas are generally used:
1. In deep water (Michell limit): H/L < 0.142.
2. In intermediate water depth (Miche formula) : H/L
< 0.14 tanh (2n d/L).
3. In the case of a solitary wave, the maximum relative
wave height: H/d < 0.78, where dis the water depth.
15-1.3 Translatory Waves
In a translatory wave, there is a transport of water in the
direction of the wave travel. Some examples of such pheno-
mena are tidal bores or moving hydraulic jumps; waves
generated by the breaking of a dam; surges on a dry bed;
undulated moving hydraulic jumps; solitary waves; and
flood waves in rivers. It is pointed out that oscillatory
and translatory waves may sometimes look very much
alike and be treated mathematically by the same method.
 Chapter 15: An Introduction to Water Waves

201

----
WHITE CAPS; LARGE WATER DEPTH.

SPILLING BREAKER: THE BOTTOM SLOPE IS GENTLE,SMALL WAVE

STEEPNESS

Figure 15-5
Different kinds of wave breakers.

PLUNGING BREAKER: BOTTOM SLOPE AND WAVE STEEPNESS ARE LARGER

SURGING BREAKER: EXTREMELY STEEP BOTTOM SLOPE.

FULLY DEVELOPED BORE IN TIDAL ESTUARY

For example, a solitary wave, which is a translatory wave, is
characterized by a unique wave crest accompanied by a
sudden jump ahead by the water particles under the wave
crest. Cnoidal waves (discussed in far greater detail below),
which are oscillatory waves, present very similar character-
istics. However, in the case of a cnoidal wave, there is a
gentle slow return under a long flat trough between wave
crests. A solitary wave motion always involves an important
net mass transport. A cnoidal wave has a very small mass
transport because of this return flow. From the mathe-
matical viewpoint, these two kinds of motion are of the
same family, i.e., they are subjected to the same simplifying
 Part 3: Water Wave Theories
202
_s_.w_.L----~ :1ft:-=-----
OSCILLATORY WAVE
SWL
TRANSLATORY SOLITARY WAVE
assumptions and they obey the same basic equations. The
solitary wave is a limited case of the cnoidal wave when the
wave period tends to infinity (see Fig. 15-6).
15-2 Mathematical Classification
15-2.1 The Significant Wave Parameters
15-2.1.1 In an Eulerian system of coordinates a surface
wave problem generally involves three unknowns: the
free surface elevation (or total water depth), the pressure
(generally known at the free surface), and the particle
velocity. Since a general method of solution is impossible, a
number of simplifying assumptions have been made which
apply to a succession of particular cases with varying
accuracy. In general, the method of solution which is used
depends upon nonlinear effects, i.e., the relative importance
of the convective inertia terms with respect to the local
inertia.
15-2.1.2 However, instead of dealing with these inertial
terms directly, it is more convenient to relate them to more
accessible parameters. Three characteristic parameters are
used. They are:
1. A typical value of the free surface elevation such as
the wave height H
2. A typical horizontal length such as the wavelength L
3. The water depth d
Figure 15-6
Difference between an oscillatory wave
and a translatory wave.
Although the relationships between the inertial terms and
these three parameters are not simple, their r~lative values
are of considerable help in classifying the water wave
theories from a mathematical viewpoint.
For example, it is easily conceived that when the free sur-
face elevation decreases, the particle velocity decreases also.
Consequently, when the wave height H tends to zero, the
convective inertia term, which is related to the square of the
particle velocity, is an infinitesimal of higher order than the
local inertia term, which is related linearly to the velocity.
In such a case, the convective inertia can be neglected and the
equations can be linearized.
Three characteristic ratios can be obtained from H, L,
and d. These are H/L, Hjd, and Ljd. The relative importance
of the convective inertia term increases as the value of these
three ratios increases.
In deep water (small Hjd, and small Ljd), the most
significant parameter is H/L which is called the wave
steepness. In shallow water the most significant parameter
is Hjd which is called the relative height. In intermediate
water depth, it will be seen that a significant parameter which
also covers the three cases is (HjL)(Ljd) 3
15-2.2 The Methods of Solution
Depending upon the problem under consideration and
the range of values of the parameters H/L, Hjd, and Ljd,
three mathematical approaches are used. They are (1) linear-

ization; (2) power series; and (3) numerical methods.
Statistical methods are also used to describe the complexity
of sea states or waves generated by wind action.
15-2.2.1 The simplest cases of water wave theories are,
of course, the linear wave theories, in which the convective
inertia terms are completely neglected. These theories are
valid when HjL, Hjd, and L/d are small, i.e., for waves of
small amplitude and small wave length in deep water. For
the first reason they are called the "small amplitude wave
theories." This is the infinitesimal wave approximation.
The linearized equations are so amenable to mathemati-
cal solutions that the linear wave theories are used for an
extreme variety of water wave motions. For example,
some phenomena which are studied by this method are
wave diffraction, waves generated by a moving ship, and
waves generated by explosions, even though they may
have large amplitudes.
15-2.2.2 Solutions can also be found as a power series in
terms of a parameter small compared to others. This small
quantity is H/L in deep water or H/d in shallow water. In
the first case (development in terms of H/L, or Stokes
waves) the first term of the power series is a solution of the
linearized equations. In the second case, the first term of the
series is already a solution of nonlinear equations; these
are the cnoidal waves.
The calculation of the successive terms of the series is
rather cumbersome so that these methods are used in a
very small number of cases. The most typical case is the two-
dimensional progressive periodic wave. In this case, the
solution is assumed to be a priori that of a steady-state
profile, i.e., a function such as F = f(x - Ct) where Cis a
constant equal to the wave velocity. Cis also called the phase
velocity.
The simplification introduced by such an assumption
is due to the fact that
oF oF
ox o(x- Ct)
Chapter 15: An Introduction to Water Waves
and 203
_a
F = c -::-:--aF-----=:-:-
ot o(x - Ct)
so that
oF -C-
oF
at ox
As a result, the time derivatives can be eliminated and
replaced by space derivatives. Power series solutions are
also found in the case of periodic standing waves and irregu-
lar waves.
15-2.2.3 It may happen that a steady-state profile does
not exist as a solution, in which case a numerical solution
where the differentials are replaced by finite differences, is
often used. This occurs for large values of H/d and Ljd,
when the nonlinear terms such as pu(oujox) are relatively
l~rge by comparison with the local inertia such as p(oujot).
This is the case of long waves in very shallow water.
Of course, a numerical method of calculation can be used
for solving a linearized system of equations. For example,
the relaxation method is used for studying small wave
agitation in a basin. Also, an analytical solution of a non-
linear system of equations can be found in some particular
cases. Hence it must be borne in mind that these three
methods and the range of application which has been given
indicate more of a trend than a general rule.
15-2.2.4 Aside from the three previous methods which
aim at a fully deterministic solution of the water wave
problem, the description of sea state generally involves the
use of random functions. The mathematical operations
which are used in this case (such as harmonic analysis)
generally imply that the water waves obey linear laws,
which are the necessary requirements for assuming that the
principle of superposition is valid. Consequently, such an
approach loses its validity for describing the sea state in
very shallow water (large values of Hjd and L/d) and in the
surf zone.
 Part 3: Water Wave Theories

204 When ('1 0 /L)(L/d) 3 ~ 1, the linear small amplitude wave

15-2.3 An Introduction to the Ursell
Parameter
15-2.3.1 The potential function for a Stokes wave or
irrotational periodic gravity wave traveling over a constant
finite depth at a second order of approximation is found to
be Equation 15-1.
The series is convergent in relatively deep water, and
the term in His the solution obtained by taking into account
the local inertia only, while the term in H 2 is the most
significant correction due to convective inertia. Therefore,
the relative importance of the convective inertia term can
be described by the ratio of the amplitudes of these two
terms. Even though this solution loses its validity in very
shallow water, it is seen after some simple calculations that
the ratio of the amplitude of the second-order term to the
amplitude of the first -order term is:
3 1 (L)d
16 (2n) 2
H
L
3
when A = md = -
2nd
L
--.
theory applies. In principle more and more terms of the
power series would be required in order to keep the same
relative accuracy as the Ursell parameter increases.
15-2.3.2 Also, in the case of very long waves in shallow
water such as flood waves, bore, and nearshore tsunami
waves, the Ursell parameter is difficult to use since the
interpretation of L is not clear. The relative amplitude H/d
is then a more significant parameter to assess the impor-
tance of the nonlinear terms. In this case, the vertical
component of inertia force is negligible and the only
significant term for convective inertia is pu(oujox). Then it is
possible to calculate the ratio of amplitude of convective
inertia to the amplitude of local inertia, pu(oujox)/p(oujot),
directly. Since in very shallow water d/L is very small,
one has simply, using only the linear term,
oc/J H k
u = - ox = - 2 md cos (mx - kt)
0 and it is found that

since cosh A--. 1 and sinh A--. A. When (H/L)(L/d) 3 is

pu ou
OX
I
max H
very small, the small amplitude wave theory is valid. oul = 2d
If, instead of H, one uses the maximum elevation '7o p-
ot max
above the still water level (17 0 is equal to H/2 in the linear
theory), the so-called Ursell parameter initially introduced which demonstrates the relative importance of the ratio
by Korteweg and de Vries is obtained: Hjd. Despite these difficulties of interpretation, the Ursell
parameter is a useful simple guide, but is not necessarily
sufficient for judging the relative importance of the non-
linear effects.

Linear, first-order term Second-order term

proportional to H proportional to H 2

"' __ H ~ cosh m(d + z) . (k _ ) _ ~ (H) 2 k cosh 2m(d + z) . 2(k _ ) (15-1)

'+'- 2m "hd smt mx 82 .h4 md sm t mx
sm m sm
 Chapter 15: An Introduction to Water Waves

205
LIMIT DEEP WATER WAVE
STOKES(4th
orde
~ =0.14

STOKES(3rd
order)
L()
0
0
STOKES(2nd
order)

DEEP
SHALLOW
N"
(.) WATER WAVE
-4-- -.wATER
Cll WAVE

-
en
......
Figure 15-7
Limits of validity for various wave theories.

L()

/UR= I ~

I -oi....J AIRY
THEORY

I I (LINEAR)

I
I
O.OOIO_L.0-:-1:-------L.!....__ _--l.._O.J...I_ _ _ _ _ _ _ _ _.J..__ ____.Il--_ _ _ _...JIO

rd2 (ft/sec2)

15-2.3.3 Figure 15-7 indicates approximately the range
of validity of the various theories. This graph has been
established for two-dimensional periodic waves such as
illustrated on Fig. 15-8, but it gives an indication for any
kind of water waves. Three corresponding values of the
Ursell parameter have been shown. The graph is limited by
a breaking criteria which implies that there is a maximum
value for the wave steepness which is a function of the
relative depth (see Section 15-1.2.7). A comprehensive
quantitative investigation of the error which is made by
using various theories in various domains has not been
done so far; hence, such a graph is somewhat arbitrary
and merely qualitative.
15-2.4 The Two Great Families of Water
Waves
In hydrodynamics the water wave theories are generally
classified in two great families. They are the "small ampli-
tude wave theories" and the "long wave theories." The
small amplitude wave theories embrace the linearized
Part 3: Water Wave Theories

(a) AIRY WAVE : DEEP WATER , SMALL WAVE STEEPNESS

(b) STOKES WAVE: DEEP WATER, LARGE WAVE STEEPNESS

Figure 15-8
A physical illustration of
various wave profiles.
~---------~ ~---------~
(C) CNOIDAL WAVE : SHALLOW WATER

(d) SOLITARY WAVE: LIMIT CURVE FOR CNOIDAL WAVE WHEN THE
PERIOD TENDS TO INFINITY.

theory for infinitesimal amplitude waves and the first
categories of power series, i.e., the power series in terms of
H/L for finite amplitude waves.
The long wave theories embrace the numerical method
of solution mostly used for the nonlinear long wave
equations.
These two great families include a number of variations
and some intermediate cases presenting some of the charac-
teristics of both families. For example, the cnoidal wave, the
solitary wave, and the monoclinal wave are considered as
being particular cases (steady-state profile) of the long
wave theories, because they are nonlinear shallow water
waves.
It can be considered that there exists some arbitrariness
in such classification. This arbitrariness is the heritage of
tradition, since the wave theories, as any theory, have been
developed in a haphazard manner. But it is most important
to understand the relative position of these theories with
respect to each other, and their limits of validity. The small
amplitude wave theories and the long wave theories are
now considered separately.
15-3 The Small Amplitude Wave
Theory
15-3.1 The Basic Assumption of the
Small Amplitude Wave Theory
It has been mentioned in the previous section that the
small amplitude wave theory is essentially a linear theory,
i.e., the nonlinear convective inertia terms are considered
small. It is called the small amplitude wave theory because
the equations are theoretically exact when the motion
tends to zero even if the convective inertia terms are taken
into account. Indeed, in that case the nonlinear terms are
infinitesimals of higher order than the linear terms.
This assumption is extremely convenient because the
free surface elevation can a priori be neglected; i.e., the
motion takes place within known boundaries. This assump-
tion is used in order to determine the zero wave motion
and such solution is assumed to be valid even if the wave
motion is different from zero.
Aside from this assumption, the motion is also most
often considered as irrotational. This assumption is

compatible with the neglect of the quadratic convective
term pV x curl V (see Section 4-4.3). Then the solution of
the problem consists of determining the velocity potential
function (x,y,~,t) satisfying the boundary conditions at
the free surface and at the limit of the container.
This approach has been proven to be extremely successful
even for wave motion of significant magnitude and in
shallow water. Moreover, the assumption of linearity
permits the determination of a complex motion by super-
position of elementary wave motions.
15-3.2 The Various Kinds of Linear
Small Amplitude Waves
15-3.2.1 Progressive periodic two-dimensional linear
wave motion is the basic motion which leads to the under-
standing of many other more complex motions. Such a
solution is found by assuming that the motion is of the form
A sin (2n:/L)(x - Ct) where Cis a constant. A general solu-
tion is obtained as a function of water depth. In deep water,
an asymptotic solution is obtained as a limiting case. It is
valid when the relative water depth: d/L is larger than 1/2.
The deep water wave theory can also be obtained
independently. In very shallow water (d/L < 1/20) the
solution is the linear shallow water wave theory, which is a
special case (a limiting case) of the nonlinear long wave
theory. As for the nonlinear long wave theory, the vertical
acceleration is negligible and the pressure distribution
tends to become hydrostatic, i.e., the pressure is equal to the
product of the specific weight of the fluid pg by the distance
from the free surface. Also, the velocity distribution is
uniform along a vertical.
Two periodic progressive waves of slightly different
period traveling in the same direction form a succession of
wave trains that give rise to a beat phenomenon. It has been
seen that two progressive periodic waves of the same period
and amplitude traveling in opposite directions form a
standing wave or clapotis, and in the case where d/ L < 1/20,
a seiche (Fig. 15-9). A periodic wave reflected by a vertical
wall at an angle forms a system of "short-crested" waves
which appear as a grid of peaks of water moving parallel
to the wall.
Chapter 15: An Introduction to Water Waves
207
FUNDAME NT4L
HAR MONIC
Figure 15-9 Seiche motion in a two-dimensional basin.
A great number of three-dimensional periodic motions
within complex boundaries can be determined by the small
amplitude wave theory. They are the three-dimensional
wave motions within tanks of various shapes (rectangular,
circular, etc.) with constant or varying depth. The process of
wave diffraction by a vertical wall or through a breach is
also thoroughly analyzed by the small amplitude wave
theory.
Finally, this theory for progressive waves is essentially
destined to be the foundation of the study of wind waves,
although this phenomenon is random and not periodic.
Hence the study of wind waves will require further analysis
as described in Appendix A on wave spectra.
15-3.2.2 The small amplitude wave theory is also partic-
ularly successful in determining the wave motion created
by a sudden disturbance or impulse at the free surface, or
at the bottom (see Fig. 15-10). For example, tsunami waves
generated by earthquakes can be treated in deep water by
mathematical methods of the small amplitude wave
theory. Likewise, waves generated by an underwater
explosion or by the drop of a stone on the free surface of a
body of water receive a similar theoretical treatment. In
 Part 3: Water Wave Theories
208
general, these waves have a cylindrical symmetry but are
nonperiodic. One may consider that they have a "pseudo"
wave period defined by the time which elapses between two
wave crests. This period decreases with time at a given
location. Also, in general, this "pseudo" wave period
tends to increase with the distance from the disturbance.
Waves generated by disturbances often appear as a succes-
sion of wave trains, the number of waves within each wave
train increasing as the distance from the disturbance
increases.
The average wave height also tends to decrease with
distance due to the double effects of the increase of wave
length with distance and radial dispersion.
15-3.2.3 Finally, the wave motion created by a moving
disturbance (ship or atmospheric disturbances) can also
be analyzed by application of the small amplitude wave
Figure 15-10
Wave generated by a
free surface local disturbance.
theory. The theoretical wave patterns created by a moving
ship are presented in Fig. 15-11. These are called Kelvin
waves.
15-3.3 The Finite Amplitude Wave
Theories
15-3.3.1 The solution for a progressive harmonic linear
wave over a horizontal bottom is a sine function of(x - Ct),
so the free surface is perfectly defined by a sine curve. In
shallow water, the crest has a tendency to become steeper
and the trough flatter, as shown in Fig. 15-8c. In this case,
the linear small amplitude wave theory is no longer valid.
15-3.3.2 In the simple case of periodic waves, either
progressive or standing, the small amplitude wave theory
can be refined by taking into account the convective inertia
forces to some extent. It has been indicated in a previous
Figure 15-ll
Wave generated by a moving ship.

section (Section 15-2.2.2) that this is done by assuming the
solutjon for the motion is given by a power series in terms
of a quantity which is small compared to the other dimen-
sions. For example, in the simple case of a periodic progres-
sive or standing two-dimensional wave, it is assumed
that the solution for the motion is given as a power series
in terms of the wave height H (or of the wave steepness
H/L, defined as the ratio of the wave height to the wave-
length L). For example, the potential function cf>(x,z,t) will
be written:
cf>(x,z,t)Hc/> 1 + H2 c/> 2 + H3c/> 3 + H4 c/> 4
=
The first-order term Hc/> 1 is found from the linear small
amplitude theory by neglecting the nonlinear terms. The
other terms are correction terms due to the nonlinear
convective inertia. These terms of the series are obtained
successively by recurrence formulas. A third-order wave
theory, known as a third-order approximation, is a theory
in which the calculation has been performed up to the
third power of the small quantity. In this case, a third-order
approximation is
Hc/>1 + H2 c/>z + H3c/>3
c/> =
In the case of a harmonic wave, c/> 2 and c/> 3 are sinusoidal
functions of n(x - Ct) where n is an integer equal to the
order of the considered term. The 4>n are functions of the
relative depth d/L. In practice the complexity of the terms
c/> 2 , c/> 3, ... increases so much as the order of approximation
increases that calculation can rarely be performed at a
high order of approximation. The formulas for the high
order of approximation are so complicated that for their
application a set of tables obtained from a high-speed
computer is required.
In engineering practice, the first-order wave theory is
most often sufficient. However, higher-order wave theory
indicates some interesting trends for waves of large steep-
ness (large H/L) in deep water. In very shallow water the
convective inertia terms are relatively high and the con-
vergence of the series becomes very slow. The series are not
even necessarily uniformly convergent and the function of
relative depth d/L loses its meaning.
Chapter 15: An Introduction to Water Waves
It has been mentioned that in shallow water the impor- 209
tant parameter becomes H/d instead of H/L for deep water.
A power series in terms of H/d is most convenient and, in
principle, would require fewer terms for a better accuracy.
Such power series appear in the cnoidal and solitary wave
theories which will be discussed in the following para-
graphs (Section 15-4.2.1 ).
15-3.3.3 Once all the equations of motion and the
boundary conditions at the free surface and at the bottom
have been specified, an infinite number of solutions may be
found. These equations are not sufficient for determining
the wave motion.
Two other conditions are required. One is on rotation-
ality and mass transport and is considered in Chapter 17.
It is also necessary to specify whether the wave motion
should be a progressive wave, a standing wave, or a wave
train. For example, in the first case, a solution for steady-
state profile has to be found such that the solution appears
as a function of (x - Ct) where C is the constant wave
velocity. In the second case, a mere addition of two periodic
gravity waves traveling in opposite directions can be used
in the linear case and for the first term of the power series.
However, higher-order terms must be found independently
by recurrence formulas, established for the specific type of
motion (progressive or standing).
15-4 The Long Wave Theory
15-4.1 The Basic Assumptions of the
Long Wave Theory: The Long Wave
Paradox
15-4.1.1 The long wave theory applies when the relative
depth is very small. Thus, the vertical acceleration can be
neglected and the path curvature is small. Consequently,
the vertical component of the motion does not influence the
pressure distribution, which is assumed to be hydrostatic.
However, contrary to the small amplitude wave theory, the
free surface is now unknown even during the first step of
the calculations. Also, the velocity distribution along a
 Part 3 : Water Wave Theories
210 vertical is assumed to be uniform, or the particle velocity is
averaged over a vertical. (As in the case of the generalized
Bernoulli equation, a correction coefficient close to unity
should be included where quadratic terms appear. This
refinement is neglected.) Because the equations are non-
linear, the number of analytical solutions is limited to a few
particular cases.
While the small amplitude wave theory consists of
finding potential function by analytical means, the long
wave theory is most often treated by numerical methods or
graphical methods and by the use of a high-speed computer.
15-4.1.2 An error, inherent in the simplifying assump-
tions, is encountered systematically in the treatment of the
nonlinear long wave theory. The velocity of the "wave
element" is an increasing function of the water depth such
as (gh) 112 Therefore, according to the theory, the wave
elements carrying the most energy should have a tendency to
catch up with the first wave elements ahead of the wave
(see Fig. 15-12). A vertical wall of water should soon result,
forming a tidal bore. However, although this phenomenon
may actually occur physically, when it does occur, it
happens much later than predicted by the long wave theory.
In particular, in the case of a wave which contains high
space derivatives for oryjox and oujox, the long wave theory
Figure 15-12 A physical illustration of the long wave
paradox.
I : 0
BORE
may no longer be valid. Similarly, the breaking of a long
wave on a beach will be predicted sooner than if it were
due to the change of bottom depth only.
Finally, it is realized that the long wave theory and the
steady-state profile are two concepts theoretically in-
compatible, although steady-state profiles have been
observed. This inherent deficiency in the long wave theory
is the long wave paradox and is also encountered in gas
dynamics and nonlinear acoustics.
The two stabilizing factors which explain the existence of
steady-state profile are the vertical acceleration and the
bottom friction. They are now considered successively.
15-4.2 Steady State Solutions
15-4.2.1 If one takes vertical acceleration into account,
the pressure distribution is no longer hydrostatic. In
particular, due to the centrifugal force of water particles
under a wave crest, the pressure at the bottom decreases
significantly (see Section 14-5.2).
Although nonnegligible, the vertical acceleration can
still be linearized by assuming dwjdt ::::; owjot. Since the
vertical component of the motion is small, the convective
terms w(owjox) and u(owjox) remain small. If this correction
effect is taken into account in the long wave theory, the
motion becomes nonlinear horizontally and linear verti-
cally. Even if the nonlinear vertical components are taken
into account in a solution obtained as a power series in
terms of H/d, they have to be introduced in high-order terms
only for the sake of consistency in the approximations. If it is
assumed that the solution of the equation oflong waves with
a correction term for flow curvature is that of steady state,
this solution has to be defined as a function of (x - Ct).
Such solutions do exist. They are the solutions for solitary
waves and cnoidal waves (although in the latter case the
pressure is found to be hydrostatic at a first order of
approximation). In these cases, the value of the Ursell
parameter does not need to be small (see Section 15-2.3.1).
In the case of very long waves where the vertical accelera-
tion and path of curvature are effectively negligible, the
choice of the significant length is quite arbitrary. This
Additional material from An Introduction to Hydrodynamics and Water Waves,
ISBN 978-3-642-85569-6, is available at http://extras.springer.com
 Chapter 15: An Introduction to Water Waves

kind of motion corresponds, however, to a value of the 2I I

Ursell parameter much larger than unity.
15-5.2 A Plan for the Study of Water
15-4.2.2 A quadratic, or more generally a nonlinear, Waves
friction from the bottom slope has a stabilizing effect which The plan in this book for the study of water waves
may balance the horizontal components to slow down follows tradition, i.e., the two families of water waves are
most of the wave elements having the highest particle separated under the titles "small amplitude wave theory"
velocity, i.e., carrying the most energy. and "long wave theory."
Under certain conditions inherent to the characteristics Chapter 16 is entirely devoted to the linear small
of the bottom slope, the friction factor, and the water depths amplitude wave theory. The case of irrotational frictionless
ahead and after the transient wave, it may happen that a harmonic motion only is considered. Two-dimensional and
steady-state translatory wave also exists. It is the mono- three-dimensional motions are studied.
clinal flood wave which is an exact solution of the long Chapter 17 deals with finite amplitude waves and Chapter
wave equation with bottom friction. 18 is devoted to the long wave theory. The theories of
solitary waves and monoclinal waves are also presented.
15-5 A Synthesis of Water Wave Appendix A is devoted to the concept of wave spectrum
Theories as an application of linear theory to phenomena of random
characteristics. Appendix B is on similitude and scale model
15-5.1 A Flow Chart for Water Waves technology, with particular emphasis on engineering
The flow chart just preceding this page summarizes the problems associated with water waves.
previous considerations. It describes the main character-
istics of water waves. The two categories of motion are the
linear and nonlinear motions, depending upon whether
the convective inertia is taken into account or not. Each
of these motions may also be subdivided into motion where
the pressure is assumed to be hydrostatic or motion with
nonnegligible flow curvature. Finally, the motion may or
may not be irrotational and the bottom friction may or
may not be taken into account.
Only the main theories have been indicated. Variations of
these main theories exist in the literature. The details of the
mathematics involved in these differences are beyond the
scope of this book. An aspect of these is presented in Chapter
17. Due to the limitation of mathematical methods, the
most complex cases cannot be analyzed. For example, a
theory for nonlinear rotational waves in shallow water
with nonnegligible vertical acceleration and quadratic
bottom friction does not exist yet.
The flow chart can be used as a guide throughout the
following chapters. However, all the theories given in the
flow chart will not be studied in detail in this book.
 Chapter 16
212
16-1 Basic Equations and Formulation
of a Surface Wave Problem
16-1.1 Notation and Continuity

Linear The motion is defined with respect to the three axes in a

Cartesian coordinates system. OX, OY, and OZ are the
mutually perpendicular axes. The OZ axis is taken to be
Small Amplitude vertical and positive upwards. Any point is defined by the
coordinates x, y, and z. The depth is defined by z = -d,
Wave Theories and is assumed to be constant (see Fig. 16-1). Viscosity
forces are neglected. The motion is assumed to be irrotation-
al and the fluid is incompressible.
curl V = 0 or ( = '1 = ~ =0
(Note that '1 in this chapter will be used for the free surface
elevation and not for vorticity.) Also,

div V = 0 or -
ou + -OV + -OW = 0
ox oy oz
These assumptions result in a number of simplifications.
curl V = 0 ensures the existence of a single-valued
velocity potential function ljJ(x,y,z,t) from which the velocity
field can be derived. Thus, the potential function can
arbitrarily be defined as V = grad or V = -grad. The
latter definition is used in this chapter, i.e., u = - ol/Jjox,
v = - o/oy, w = - o/oz. The velocity potential function

y
z

Figure 16-1
Notation.
 Chapter 16: Linear Small Amplitude Wave Theories

has to be found from the continuity equation, the momen- 213

tum equation and the boundary conditions.
The continuity equation div V = 0 is expressed in terms
of by the equation '\1 2 = 0. In Cartesian coordinates,
it is written as
16-1.3 Boundary Conditions
16-1.3.1 At a fixed boundary, the fluid velocity is tangen-
tial to the boundary, that is, the normal component V, is
zero. In terms of velocity potential , this condition IS
written o/on = 0. In particular on a horizontal bottom

16-1.2 The Momentum Equation w I --- o1 -0

z= -d - OZ z= -d -

The momentum equation for an irrotational flow is given

by the following form of the Bernoulli equation (see Section 16-1.3.2 One of the difficulties encountered in determin-
10-1.2). The minus sign in Equation 16-1 is due to the new ing the nature of wave motion is due to the fact that one of
definition of. the boundaries-the free surface-is unknown, except in
the case of infinitely small motion in which the free surface
o
+ ! vz +
p
- + gz = f(t) (16-1) is, at the beginning, assumed to be a horizontal line. Hence,
ot 2 p
another unknown z = 11 appears in wave problems. If one
Local Convective Pressure Gravity
assumes that the free surface, in the most general case of a
inertia inertia term term
term term three dimensional motion, is given by the equation
z = rt(x,y,t), then the variation of z with respect to time t is
In this equation, f(t) may depend on t but not on the space
variables. The fact that the flow is assumed to be irrota- dz 011 011 dx 011 dy
tional means that the Bernoulli law is valid throughout the -=-+--+--
dt ot ox dt oy dt
fluid and not only along streamlines.
This equation is nonlinear because of the convective Introducing the values
inertia term. This term may be expressed as a function of the
potential function so that dx o dy o
-=U= - - -
dt-- v-
-
dt ox oy

The nonlinearity of the motion can be seen clearly. In the

case of very slow motion, the convective term is neglected
and the Bernoulli equation is written as
o p
- - + - + gz = f(t)
ot P
Periodic gravity wave theories often satisfy the condition
for slow motion with a fairly good degree of accuracy. The
corresponding solutions are mathematically exact when
the motion tends to be infinitely small.
the free surface equation becomes
This equation is nonlinear and is the kinematic condition
at the free surface.
Another equation-the dynamic equation-is given by
the Bernoulli equation in which the pressure p is considered
as constant (and equal to atmospheric pressure). Hence the
 Part 3: Water Wave Theories

214 free surface dynamic condition becomes z

8ot 2 [(8)
- - + -1 -
2+ (8)2
- + (8)2]
- + g1] = f(t)
ox oy oz
Thus, generally fjJ and 17 appear to be given by the
solution of V2 = 0 with two simultaneous nonlinear
boundary conditions at the free surface and a linear
boundary condition at the bottom

81
OZ z=-d-
-0
Figure 16-2 Notation.

16-1.4 The Free Surface Condition in the velocity at the free surface is now equal to the normal
Case of Very Slow Motion velocity of the surface itself. This gives with sufficient
In the case of slow motion the Bernoulli equation approximation
8
- - + - + gz =
p
f(t) ~~ =- ~~Lo
ot P
17 may now be easily eliminated from the dynamic and
becomes kinematic conditions. The derivative of 17 with respect to t
- 81
ot z=~
+ g1] = 0
in the dynamic condition gives

at the free surface. The result is: ~~ 1 ~:~~z=O

=

1]-! 8 I
- g ot z=~
01J/8t can be eliminated by equating the two above equa-
tions. This yields :
provided the function f(t) and any additive constant can
be included in the value of of/Jjot. [~~ + ~. ~:~1=0 = 0
Since the motion is assumed to be infinitely small, 17 which is called the Cauchy-Poisson condition at the free
may be written surface.

~ ~; lz=O 16-1.5 Formulation of a Surface Wave

Problem
This approximation leads to an error of the order of those 16-1.5.1 Thus fjJ and 17 appear to be solutions of the
already done in neglecting the convective inertia term. following system:
Consider the kinematic condition: 81]/ox and 817/8y are 1. Continuity
the components of the slope of the free surface and are
- d ::;; Z ::;; 1](X,y,t)
small as in the case of slow motion (see Fig. 16-2).
{
The nonlinear terms (of/Jjox)(o1J/ox) and (8joy)(81J/8y) -oo<x<oo}
no boundary
may be neglected. The normal component of the fluid -oo<y<oo
 Chapter 16: Linear Small Amplitude Wave Theories

2. Fixed boundary ajan = 0. In particular, at the 215

bottom 16-2 Method of Solutions
a1 - o
az z= -d
16-2.1 General Approach
16-2.1.1 When all the equations are homogeneous and
3. Free surface z = 17(x,y,t) linear, the principle of superposition states that any number
a. Kinematic condition: of individual solutions may be superimposed to form new
functions which constitute solutions themselves. In a
linear equation and its derivatives occur only in the first
degree in every term. For example, if 1 and 2 are two
separate solutions, a 1 + b 2 is also a solution, a and b
b. Dynamic condition:
being two arbitrary constants. This basic principle is very
[(a) + (a) 2+ (a)
important and will be used in the following sections.
_ a +! 2 2
] + g 11 = 0
at 2 ax ay az
16-2.1.2 Most of the solutions with which we are con-
where f(t) is now included in a/at. However, cerned in this chapter are harmonic. This stems from the
even in this case, this last equation may be different fact that harmonic functions are quite natural solutions of
from zero, and equal to a given function f(x,y,t) the basic equations. The solutions characterizing periodic
in the case of a disturbance created at the free motions may be considered as superposition of harmonic
surface. components.
So formulated, the solution of the system of equations The solution of (x,y,z,t) is usually of the form
presented in this section is still difficult to determine.
First, the equations are nonlinear, and second, the free = f(x,y,z) cos (kt + t:),
surface is unknown and is time-dependent. where k = 2n/T and T is the wave period. Another form
of the solution is
16-1.5.2 In the case of slow motion, 17 may be eliminated
from the two free surface conditions resulting in the simple = Re f(x,y,z)ei<kt+e)
Cauchy-Poisson condition Recall that

[a2 + g a] - 0 ei(kt+e) = cos (kt + t:) + i sin (kt + t:)

at 2 az z=O -
"Re" means the real part of the fi.mction and t: is the
This leaves only one unknown, , to be determined from phase of with respect to the origin of time, t = 0. In the
following, "Re" will be omitted and it is to be understood
-d::::; z::::; 11 = 0 that only the real parts of the mathematical expressions are
{ -oo<x<oo considered.
-oo<y<oo Introducing this form of in the free surface condition

o1
-
OZ z=-d
= 0 and [a-Ot+ a]
2
2
OZ
g-
z=O
=0
 Part 3: Water Wave Theories

216 gives The equations

( k2 - _E_) = 0 P"
g oz p

16-2.1.3 If it is assumed that is given by a product of are now reduced to

functions of each variable alone, then the basic equation
V2 = 0 may be solved by the separation of variables
method. From physical considerations, it may be expected
that the solution will be given by the product of the func-
tions of the horizontal components U(x,y), the vertical and
component P(z), and the time f(t). Hence oU oU
2 2
ox2 + oy2 + m2U(x,y) = 0
= U(x,y) P(z) f(t).
This value of cp(x,y,z,t) can be used in the continuity These equations will often be written in the shorter form:
equation, V2 = 0. Algebraic manipulation of the result

{ (~2 - )P
will give: 2
m2 = 0
o2U/ox 2 + o2Ujoy 2 d2 P/dz 2
U(x,y) P(z) (V 2 + m2 )U = 0
This may be written as The last equation is the well-known Helmholtz equation
(also called the wave equation) of mathematical physics.
V2 U P"
u p
16-2.2 Wave Motion along a Vertical
Notice that the functions of x and y are on one side of the
equal sign, while the functions of z are on the other. The The equation
variables have been separated.
It must be said that it was not certain at the beginning
that it would have been possible to separate the variables as
has been done. However, it will be shown later that this may easily be integrated, giving the general solution
process may be performed for solutions of V2 = 0. The
right-hand side of the above equation is a function of z.
The left-hand side is a function of x and y. Since x and y
where A and B are constants.
can vary independently of z and vice versa, the only way in
The boundary condition at the bottom,
which the function of x and y and the function of z can
always be equal (as stated by the above equation) is if the
left-hand side and the right-hand side are both equal to the 81
OZ
-0
z= - d -
same constant m 2 where m may be real or imaginary.
It will be easily seen that if m is imaginary there is no gives for any fixed value of x, y, and t:
physical meaning to the solutions in the case of wave
motion. Thus, m is chosen to be real and m2 is always dPI - o
positive. dz z= -d
 Chapter 16: Linear Small Amplitude Wave Theories

When this boundary condition is applied to the solution where a, f3 are constant coefficients which depend upon 217
for P the result is the boundary conditions, k is the frequency 2n/T, and T is
mAe-md- mBe+md = 0 the wave period. When f3 = 0 and the coefficient rx is in-
cluded in the coefficient D, it is found
Hence
cfJ = D cosh m(d + z)V(x,y)eikt
Since there exist an infinite (but discrete) number of values
Consider the original solution, for kn and mn which satisfy the equation
p = Aemz + Be-mz k?; = mng tanh mn d
Multiply each term on the right by e-mde+md. Let a general solution for cjJ can be written as
Ae-md = Be+md = !D 00

cjJ = L Dn cosh mn(d + z)Un(x,y) exp (iknt +en)

and substitute this in the equation for P. The result is n=O

where en is a phase constant.

p = !!_ (em<z+d) + e-m(z+d)) Consider the case of a monochromatic wave. It is
2
convenient to express D as a function of the wave height 2a.
That is P = D cosh m(z + d). From the free surface dynamic equation for slow motion,
Now, substituting P in the expression of cjJ gives
I ] -1-81
-
cjJ = D cosh m(d + z)V(x,y)f(t) g 8t z=O
one obtains
16-2.3 Introduction of the Free Surface
ikD .k
Condition: General Solution I]=- cosh mdU(x,y)e' 1
g
The solution for f(t) is given by the Cauchy-Poisson
condition at the free surface: Considering only the real part, this can be written as
(recall i2 = -1):
( 8z2 + g 8) - o 1J = - kD cosh mdU(x,y) sin kt
8t 8z z=O -
g
Substitute the value of cjJ obtained in the previous section The expressions for cjJ and 1J become more convenient if
in this equation. Only the case when z = 0 needs to be
we write (aU) for the amplitude of IJ. Then
considered here, since this is the free surface condition.
When the resulting equation is divided by the value of , kD
a=- -coshmd
the following is obtained: g
Hence
!"If = - gm tanh md ag 1
If we let k 2 = gm tanh md, the solution for f is given by D=
k cosh md
the equation f" + k 2f = 0. The characteristic equation and
r2 + k 2 = 0 gives r = ik. Hence
A.. = _ ag cosh m(d + z) U( ) ikt
f = aeikt + {3e- ikt .,., k cosh m d x,y e
 Part 3: Water Wave Theories

218 Substituting the relationship k 2 = mg tanh md leads to This is the velocity potential function of a progressive wave
traveling in the OX direction.
,~,. _ _ !!:._cosh m(d + z) ( ) ikt If U = eimx, the velocity potential function of a wave
'f' - a . h d U x,y e
m sm m traveling in the opposite direction is obtained.
With the value of D that has been found, the expression for If the solution for U is:
P becomes
ak cosh m(d + z) ag cosh m(d + z)
P(z) =
m sinh md k cosh md or
Under these conditions the wave height at any point is 1 . .
2aU(x,y). U(x,y) is the relative value of the wave height with u= 2i (e'mx- e-mx) = sin mx
respect to a plane or a point where it is simply 2a.
then
16-3 Two-Dimensional Wave Motion ,~,. _ _ !!:._cosh m(d + z) {cos} k
'f' - a . h d . mx cos t
m sm m sm
16-3.1 Integration of the Wave Equation This is the velocity potential function of a standing wave.
The differential equation to be solved is (V 2 + m 2 )U = 0. If A' is different from B', a partial standing wave is obtained.
A general solution of this equation does not exist, but a In practice, the values for A' and B' are given by vertical
number of solutions may be found, corresponding to boundary conditions (wave reflection, etc.).
particular boundary conditions. In the case of a two- In the most general case of a two-dimensional irregular
dimensional wave such as motion encountered in a wave wave, as may be observed at sea, the velocity potential
flume. function is:
kn cosh mn(d + z) .
L
00

l/J = -an- . h exp [1(knt- mnX +en)]

n=O mn Sill mn d
This reduces the wave equation to where en is a phase constant.

(..!:.._
ox 2 + m
2 )u = 0
When there are two waves only, traveling in the same
direction, the velocity potential function describing the
"beating" phenomena may be obtained easily.
Solving this, one finds that the solutions for U are given by
any linear combination of e-imx and eimx such as,
16-3.2 Physical Meaning: Wavelength
U= A'eimx + B'e-imx
It is easy to see the physical meaning of the coefficient
In particular if U = e-imx, then m. Since and consequently 11 is periodic with respect to
k cosh m(d + z) i(kt-mx) space, m = 2n/L and Lis the wavelength.
c/J= -a-
m sinh md
e The wavelength is given by
or
k2 = mg tanh md
= -a!!:._ cos~ m(d + z) cos (kt - mx)
m smh md and then
 Chapter 16: Linear Small Amplitude Wave Theories

(-2n)
T
2 2n
L
2n
= - g tanh-d
L x = x0 - a
cosh m(d + z0 )
. h d
SID m
cos (kt - mx 0 )
219

that is, and

sinh m(d + z 0 ) .
gT 2 2nd z = z0 +a . h d
SID m
SID (kt - mx 0 )
L =-tanh-
2n L
Squaring and adding these two last equations to eliminate
and the wave celerity: t, the equation of an ellipse is obtained
L gT 2nd (x - x 0 ) 2 (z - z0 ) 2 _ 1
C=-=-tanh-
T 2n L
Az + B2 -
It is now seen that x 0 and z 0 are at the center of the ellipse,
In particular when d/L is small (shallow water) i.e., can be considered as the position of the particle at rest
__. 2nd with the horizontal semimajor axis
tanh 2nd
L :.:: L L = T(gd)112 c= (gd)l/2
+ z0 )
A = a cosh m(d
sinh md
and
and the vertical semiminor axis:
L = T(gd)t;z c = (gd)l/2 B = a sinh m(d + z 0 )
When d/L is large (deep water), tanh 2nd/L = 1, and sinh md
L = gT 2 j2n, C = gTj2n. The values of Land Care given B = a at the free surface, and B = 0 at the bottom (Fig.
as functions of the depth d and the wave period T on the 16-6). The free surface equation is
following nomographs (Figs. 16-3, 16-4, and 16-5).
1 a .
= - - = a SID (kt - mx)
g at
1J
16-3.3 Flow Patterns
When d -H:fJ, (A/B) ~ 1 and the orbits are circles of radius
The velocity components are u = - ajax, w = - ajaz R = a exp (4n 2 z/gT 2 ).
and the particle orbits are: In the case of a standing wave, it would be easily found
that the paths of particles are straight lines given by
X= - ft acp dt z = - ft a dt z - z0
0 ax 0 az -tanh m(d + z0 ) cot mx 0
X - Xo
In the case of a progressive wave: or
z - z0
cosh m(d + z) . (k ) - - - = tanh m(d + z0 ) tan mx 0
u = ka SID t- mx X - Xo
sinh md
(see Fig. 16-7). They are parabolas at a second order of
sinh m(d + z) approximation.
w = ka . h md cos (kt - mx)
SID
16-3.4 Partial Standing Wave
The particle orbits are determined by assuming that the A partial standing wave is caused by the superimposition
motion around a fixed point x 0 , z 0 is small, so that one can of two waves of the same period but travelling in opposite
consider x and z constant in the integration. directions and with different amplitudes (Fig. 16-8).
 Part 3: Water Wave Theories

220

10

20

X>
40
601
.....
..
u

0
Figure 16-3
0
Wavelength vs depth and period.
~ 6 ~~-#~~~~~~~~~---+----~----+-~~
0..
vr 2
L : ---
2Tr

vT 2 2Trd
L =- t a nh -
2Tr L

0 ~--~----~----~----L---~-----L----~--~
0 toO 200 300 400ft
10 ~ ~ ~ ~
WAVE
60LENGTH
ro ~ ~ 100 oo ~~ ...

The first-order potential function is wave height of the two progressive waves by measuring the
k cosh m(d + z) amplitude at the antinode (maximum) A and at the node B
<P= m sinh md (minimum). Then,
x [a 1 sin (kt + mx) + a 2 sin (kt - mx)]
The amplitude at the antinode is (a 1 + a 2 )
and at the
node is (a 1 - a 2 ). It is possible to determine the individual
 Chapter 16: Linear Small Amplitude Wave Theories

60
221

55

~,,a-::;:.
50 /16
1---''4

~~
v'2-

v
45 ~ Vt'-
f--10

~ ~v-
---
f--9-
40 ~
f-B-
.,
u

~ 35 ~ ~~B;-
c,:j;j-
7

~~ ----
Cl
.,.:
~ __- ___.v -s::
/ Q
!::: a:
w

~----
u 0..
0 30
Figure 16-4 .J
w
Wave velocity vs depth and period.
>
w
_,......... ~:;':
5-
~ 25
~c' 1!...
~~~ / 2,.
~

20 4-
/
1~ L
~

f
15 3

~
10

5
r;
r;
2-

0
0 10 20 30 40 50 60 70 BO 90 100
DEPTH. ft

The reflection coefficient of the obstacle causing the partial 16-3.5 The Use of Complex Number
standing wave is: Notation
To show the great simplicity introduced by the operation
A-B W = <P + il/l, W is calculated to a first order of approxima-
R=--
A+B tion in the case of a two-dimensional monochromatic
progressive wave motion. First of all, it is convenient to
The envelopes of the crest and the trough are two sinusoids make a change in the origin of the vertical axis, and to take
of amplitude, (A - B)/2. the horizontal axis on the bottom, such as: z' = d + z.
 Part 3: Water Wave Theories

222

...
~

X
1-
(!)
Figure 16-5
z
L1.l Wave length vs period and depth .
...J

L1.l
>
<(
:I: 16 15
100

o~--...JI0~--2~o~~~--~~~~s~o~~so~--~7o~~a~o--~9~0--~IO~O---I~I0---...~12-0~
0~~--~--~--+-~
WATER DEPTH, meters

With this new condition the velocity potential function The stream function l/1 is given by one of the following
could be written operations:
k cosh mz' ot/J ol/f or
ot/J ol/f
tjJ = -a- . cos (mx- kt) W=
ox
m smhmd
U=
ox oz' oz'
If This gives
A= -a-~~~
k1 l/1 = A sinh mz' sin (mx - kt)
m sinh md Now
then w= tjJ + il/l
tjJ = A cosh mz' cos (mx - kt). = A[cosh mz' cos (mx - kt) + i sinh mz' sin (mx - kt)]
 Chapter 16: Linear Small Amplitude Wave Theories

The velocity potential function is given by the real part 223

of W, while the stream function ljJ is given by the imaginary
part.
Similarly, the value of W for a clapotis is
W = 2A sin mZ cos kt or 2A cos mZ cos kt

16-4 Three-Dimensional Wave

Figure 16-6 Two-dimensional progressive
Motion
wave motion.
16-4.1 Three-Dimensional Wave Motion
in a Rectangular Tank
The following relationships hold: The separation of variables method is used to solve the
equation governing three-dimensional motion,
cosh mz' = cos imz'
sinh mz' = - i sin imz'
Using this, one obtains
It is found that
W =A[cos imz' cos (mx - kt) + sin imz' sin (mx - kt)]

This can be rewritten as

and
W =A cos (mx- imz'- kt)
azu
Introducing the complex number Z = x - iz', one obtains ayz + qzu = 0
the very simple relationship
in which p and q are constants, real or complex, such that
W = + iljf =A cos (mZ- kt) mz = pz + qz.

I~ LI 2
'z

Figure 16-7
Standing wave.
 Part 3: Water Wave Theories

224

Figure 16-8
Variation in amplitude in a
partial standing wave.

As an application of three-dimensional motion, the It is interesting to note that the equation of a two-di-
case of a standing wave in a rectangular tank is analyzed. mensional motion,

)u
The boundaries are shown in Fig. 16-9.
+ m2
-ocp = 0 or
oU O
- =
"
10r
{X = 0, =a
X
( : :2 = 0

on on y = 0, y = b shows that the curvature of the free surface is proportional

to the free surface elevation 17 since 17 = aU. Similarly, in
It is easily verified that the solution of (V 2 + m 2 )U = 0 the case of three-dimensional motion, one has
is U = cos (rnxja) cos (snyjb) with p = rn/a, q = sn/b,
where rand s are integers. The solution for V2 c/J = 0 is
( : :2 + ::2 + m 2 )u = 0
cjJ = -a 15:._ cosh m(d + z) cos rnx cos sny Then it is the sum of the curvatures in two directions (OX
m sinh md a b
and 0 Y) which is proportional to the free surface elevation.
The general solution is This is the case, for example, of the "short-crested wave."

,~., ~ ~ ~ kn cosh mid + z) r;nx sjny

'+' = L.. L, L.. -an- . COS-- COS--
i=O j=O n=O mn Sinh mn d a b 16-4.2 Cylindrical Wave Motion
where The continuity equation expressed in terms of cylindrical
coordinates is
and
o2 cjJ o2 cjJ o2 cjJ
1 oc/J 1
or + -;: 8r + o(JZ + oz 2 = 0
2 r2
Figure 16-9 Rectangular basin-notation. In the case of a motion with an axis of symmetry so that
y o2 cpjo8 2 = 0, the solution is
cjJ = U(r)P(z)eikt
The resulting wave equation (V 2 + m 2 )U = 0 for cylin-
drical waves becomes

o +-;:or+
(arz 1 a
2
)
mz u = 0
 Chapter 16: Linear Small Amplitude Wave Theories

Solutions of peculiar interest are given by the Hankel 225

functions Hb2 >(mr) and Hb1 >(mr), which when combined with
e +ikt represent progressive (diverging or converging) waves.
Table 16-1 gives the relationships between J 0 , Y0 , and the
Hankel functions Hb2 >(mr) and Hb2 >(mr) as well as their
physical meanings and their analogy with sinusoidal
functions. The appropriate solution and its physical
significance is obtained by combining the solution with
e- ikt or e + ikt and examining its asymptotic form at large
distance from the origin. The asymptotic expressions enable
us to identify outgoing waves from incoming waves. When
- .5U-------~------~----~
mr becomes very large, it is found that
0 5 10 15
Figure 16-10 Bessel functions offirst order. Hb1 >(mr) ~ AeiB Hb2 >(mr) ~ Ae-iB
J 0 (mr) ~ A cos B Y0 (mr) ~ A sin B
This is the Bessel equation of order zero. Its standard where
solutions are denoted by J 0 (mr) and Y0 (mr), the first being
regular at the origin, the second singular at the origin. The A = - 2- and n]
B = [mr - -4
nmr 112
variation of J 0 and Y0 as functions of x = mr are given by
Fig. 16-10. Any linear combination of these standard In a two-dimensional motion, A is replaced by a and B is
solutions is itself obviously a solution of the equation. replaced by mx. It is seen that the wave height, which is

Table 16-1 Analogy between Bessel functions and sinusoidal functions

Two-dimensional waves (sinusoidal functions) Circular waves (Bessel functions)

( : :2 + m2 )u = 0 (-az2 + --
8r
1a
r ar
+ m2) U = 0

e-imx =cos mx- i sin mx Progressive wave in the positive Converging wave (sink) (wave
OX direction (wave height height increases when r
=constant) decreases) when combined with
e+ikt
e+imx = cos mx + i sin mx Progressive wave in the negative Diverging wave (source) (wave
OX direction (wave height height decreases when r
= constant) increases) when combined with
e+ikt
eimx + e-imx Standing wave (horizontal H~1 > + H~> Standing circular wave (horizontal
cosmx=---- velocity = 0 when x = 0) 1 0 (mr)= 2
2 velocity = 0 when r = 0)
eimx- e-imx Standing wave (horizontal H~1>- H~z> Standing circular wave (horizontal
sinmx=---- Y0 (mr) =
2i velocity maximum for x = 0) 2i velocity infinite for r = 0)
 Part 3: Water Wave Theories

226 given by the amplitude of U, is proportional to A and This formula is general and can be applied by inserting the
decreases as r 112 This may also be demonstrated by con- value of the potential function for any kind ofirrotational
servation of the transmitted wave energy (see Section 16-5). wave, linear or nonlinear. In the case of a linear periodic
progressive wave (see Section 16-3.1)
16-4.3 Wave Agitation in a Circular
Tank = -a}!_ cos~ m(d + z) cos (kt- mx)
The boundary condition for a circular tank is m smhmd
Inserting this value in the above expression leads to
a1 -0
ar r=R- _ p ~ Jt+T J~ 2 k 3cosh 2 m(d + z)
or Fav- a . 2
T 1 -d m smh md
x sin 2 (kt - mx) dz dt
where R is the radius of the tank. The solution tor U is After integration and the neglection of some high-order
U = Jn(mr) cos ne, where Jn is the Bessel function of order terms, this becomes
n. When n = 0, U = J 0 (mr) and the motion is a stationary 1 2 2md)
4pga C( 1 + sinh 2md
circular wave. Fav =

This tends toward

16-5 Energy Flux and Group Velocity
16-5.1 Power and Energy Flux
in deep water, and
It is seen in Section 10-3.1.1 that the power transmitted
by a fluid through a penstoke of cross section A, is given by Fav Id-o = !pga 2 (gh) 112

v2 P } AVK in shallow water.

P= {-+-+z
2g pg
Note that K = 5 0 if Pis in horsepower and A is in square
feet. The same formula applies to water waves, but due to
time dependency the wave power is obtained by averaging
with respect to time. It is then called energy flux. The
16-5.2 Energy per Wavelength and Rate
of Energy Propagation
The stored energy per wavelength per unit length of
crest is the sum of the kinetic energy and the potential
energy:

f+L
average energy flux per unit of wave crest through a fixed
vertical plane parallel to the wave crest is E= p f}!V 2 + gz) dz dx
1 ft+ Tfq (P TV2 + P + pgz )
Fav =T 1
-d U dz dt By inserting the value of V 2 = [(acpjax? + (acpjazf] and
neglecting some high-order terms, it is found that in the
i.e., by application of the Bernoulli equation where f(t) is case of a simple linear periodic progressive wave that the
assumed to be taken into account in a/at: potential energy EP equals the kinetic energy EK and the

F = -
1
p-
Jt+T Jq a a
- - dz dt
total energy per wavelength per unit length of crest is
av T t -d at ax E = EP + EK = !pga 2 L
 Chapter 16: Linear Small Amplitude Wave Theories

Per unit horizontal area, this is Since m = 2n/L, dm = -2n dL/L 2 , and 227
dC
U = C- L dL
Dividing the energy flux by the stored energy gives the When the values C = (gT/2n) tanh (2nd/L) and L = CT
rate of propagation of energy: are inserted this becomes

U E = F av =
Eav
~ (1 +
2
2md )
sinh 2md U =
c(
2
2md )
1 + sinh 2md

i.e., U E = -!(gT/2n) in deep water and U E = (gd) 1 12 in
shallow water, which means that the rate of transmitted
energy is half of the wave velocity in deep water and equal
to the wave velocity in shallow water.
16-5.3 Group Velocity
Consider the linear superimposition of two progressive
waves of the same amplitude and slightly different periods
such as
Yf = a sin [mx - kt] + a sin [(m + bm)x - (k + 6k)t]
where bm and 6k are supposed to be small quantities. This
expression can be written as:
It is interesting to note that this expression is the same
as the expression for the rate of energy propagation UE. This
stems from the fact that in the case of a wave train, there is
no energy passing through the node of the wave train where
the wave amplitude is zero. So, the velocity of energy propa-
gation is that of a wave train or of a group of waves. For
this reason, the group velocity U is equal to the rate of
energy propagation. However, it is important to note that
this statement holds true only in the case of a linear wave.
The values of U/C 0 and C/C 0 where C0 is the deep water
wave celerity (C 0 = gTj2n) is shown on Fig. 16-11. The
energy flux can now be expressed as a function of the group
velocity U:

Yf = 2a cos -!{6mx - 6kt) sin [ ( m + 6; )x -(k + 62k}J 16-5.4 Physical Interpretation

Since 6m and bk are small:
Yf ~ 2a cos !(6mx - 6kt) sin (mx - kt)
It is seen that the wave 2a sin (mx - kt) is modulated by
the term:
A physical interpretation of the wave energy flux can be
given. In deep water, the particle velocity describes a circle
as seen in Section 16-3.3. The particle velocity remains the
same in absolute value along the circle. Consider a vertical
plane AB passing through the circle (Fig. 16-12). The kinetic
energy flux in this plane is zero. This energy, p V 2 /2, is the
same at point A and at point B. The only difference is that the
directiOn is opposite. The potential energy is different at A

cos -!{6mx - 6kt) = cod bm(x - :~ t)
which travels at a speed U = 6kj6m. The wave velocity
C = k/m, and 6k = 6(mC), U = 6(mC)j6m. If these values
are taken at the limit when 6m tends toward dm, U can be
written as U = d(mC)/dm. This is just U = C + m(dC/dm).
from its value at B, and only the potential energy is trans-
mitted through the plane AB. Since the average value of
potential and kinetic energy are equal, there are only two
possibilities. Either one-half the wave energy is transmitted
with phase velocity or the total energy is transmitted at
one-half the phase velocity, which is the deepwater group
velocity.
 Part 3: Water Wave Theories

v--
1.0
228
1-"'

L_

v
__L

0.8 / Co

I
-------
u
/Co

f/
0.6
v
Figure I 6- I I
Relative wave velocity and group

(
0.4 velocity vs depth.

0.2
I

0.1 0.2 0.3 04 0.5

Figure I6-12 The potential energy only is 16-6 Wave Transformation

transmitted by the wave.
WAVE TRAVEL
16-6.1 Wave Shoaling
In the case of a wave traveling over a very gentle slope, it
is assumed that the wave motion is almost the same as if
the bottom were horizontal. In a word, the flow pattern
v deformation due to the bottom slope is neglected. It is then
assumed that the flux of transmitted energy is a constant. In
this case the wave height 2a at a given depth dis known as a
function of the deep water wave height 2a 0 from the formula
 Chapter 16: Linear Small Amplitude Wave Theories

Thus, 229

1
4pga k tanh md[ 1 + sinh2md2md
2g J 1
= 4pgao
2g
k
16-6.2 Bottom Friction and Wave Damping
At a given location such as x = 0 the velocity at the bed is
and
ub
81
= ~ .
ka
h d cos kt
2a H 1 uX z= -d Sin m
2a 0 = H0 = {tanhmd[1 + (2md/sinh2md)]} 1i 2 The friction stress r for a turbulent boundary layer is
given by
= (~0 y 12 = K 5 r = pfubiubi
where .f is the friction factor (f ~ 0.015). The average work
where Ks is the shoaling coefficient. Inserting the expression done by this friction stress per unit area or as rate of energy
for LjL 0 = tanh (2nd/L) where L 0 is the deep water wave dissipation is
length L 0 = gT 2 j2n, it is then possible to calculate the
value of H / H 0 as a function of d/ L 0 only. The results of such
calculation are shown in Fig. 16-13.

Figure 16-13 Relative wave height variation with depth.

1.4

\
1.3

\
1.2

\
1.1
0

--------
J:
......
J:
1.0

~r--
0.9

0.8
0 0.1 0.2 0.4 0.5
d/ Lo
 Part 3: Water Wave Theories

230 Since H = 2a and k = 2n/T then, between wave orthogonals. The refraction coefficient K, is
(b 0 /b) 112 , where b0 is the distance between wave orthogonals
DJ = pf -T
1 fT (nH)3
. h3 d (cos kt) 2 [cos ktl dt in deep water. The combined effects of shoaling, damping,
3
0 T sm m and refraction are obtained by multiplying the shoaling
i.e., coefficient K., the damping coefficient K J and the refraction
4 z H3 coefficient K, so that
DJ = 3Pfn T3 sinh3 md
H = KsK 1 K,H 0
The variation of energy flux equals the rate of energy
Wave refraction diagrams giving distance between wave
dissipation, i.e.,
orthogonals are obtained by graphical methods or by
computer. Refraction effects remain small as long as the
water depth is larger than L/3. Therefore, wave refraction
or, H/H 0 = KsK 1 . The damping factor K 1 is calculations (and scale models) need to be done only over
bottom topography shallower than such a depth.
KJ = (1- ~ pg
Jd(x)DJ
-oo
dx) 16-6.3.2 The first method for drawing a wave refraction
diagram consists of calculating the wave crests first (Fig.
which is solved numerically over complex topography.
16-14). The wave crests are obtained by drawing the
envelope of circles from a preceding wave crest, the radius
16-6.3 Wave Refraction
of the circles being proportional to the local values of the
16-6.3.1 A wave orthogonal is the locus of points which wavelength, or one of its multiples.
define the minimum time of travel for wave propagation The second method consists of drawing the orthogonals
between two points. Wave orthogonals are usually normal directly by application of the Snell's law ofwave refraction:
to wave crests. They indicate approximately the direction
of propagation of the wave energy. When the wave arrives Cz sin IXz = cl sin IXt
at an angle with the bottom contours and refracts, the where the subscripts 1 and 2 correspond to two successive
distance between wave orthogonals b varies by a factor of values of the water depth.
b2 cos IXz
Not only is it possible to calculate the refraction co-
efficient from the relative distance of two wave orthogonals
bl COS IX1
but it is also possible to determine it from the curvature
IX is the angle of the bottom contours with the wave crest. of the orthogonals.
Subscripts 1 and 2 refer to the two referenced water depths
on both sides of the bottom contour. When the variation 16-6.3.3 In many cases refraction diagrams provide a
of wave height along a wave crest is small (which is not the reasonably accurate measure of the changes waves undergo
case of wave diffraction) it is generally assumed that the on approaching a coast. Quite often they provide the only
wave energy flux between wave orthogonals is constant, measure of these changes available. However, the accuracy
i.e., Favb = constant, and of data determined from refraction diagrams is limited
by the validity of the theory of their construction and the
tpgH 2 Ub = constant
accuracy of the depth data on which they are based. The
Hence, the wave height varies with a refraction coefficient orthogonal direction change is derived for the simplest
equal to the square root of the inverse ratio of the distance case of straight parallel contours and, although little error
 Chapter 16: Linear Small Amplitude Wave Theories

231
SCALE

Figure 16-14 Wave refraction diagrams showing wave crests, wave orthogonals,
and relative wave heights along wave crests (shaded).

is introduced by bringing orthogonals over relatively
simple hydrography, it is difficult to carry an orthogonal
accurately on shore over complex bottom features. More-
over, the equation is derived for small waves moving over
relatively flat slopes. Although no strict limits have been set,
accuracy as far as height changes are concerned cllnnot be
expected where bottom slopes are steeper than 1/ 10.
Model tests have indicated that direction changes occur
nearly as predicted even over a vertical discontinuity.
Nevertheless, a small wave reflection takes place. A third
limitation is inherent in the assumption that no energy
travels laterally along a wave crest. No strict limits have
been set, but the accuracy of wave heights derived from
orthogonals that bend sharply is questionable. A further
problem is the interpretation of crossed orthogonals. Such
cases would seem to lead to the formation of high, short-
crested waves with breakers, and caustic, doubling wave
crests due to nonlinear effects.
In short, refraction coefficients which are quite different
from unity, say K, < 0.5 and K, > 2, must be doubted.
 Part 3: Water Wave Theories

232 When they do occur as a result of one of the plotting pro- Inserting the expression for u (at a first order of approxima-
cedures mentioned, other effects such as diffraction may be tion), yields
occurring. Nonlinear effects also influence wave diffraction,
as can be seen by the doubling of wave crests at the ends of -p IT f~ (H- k cosh m(d + z) sm. (k t - mx))z dt dz
T 0 d 2 sinh md
breakwaters.
_ 1
- u;pgH
2[
1
4nd/L
+ sinh (4nd/L)
J
16-7 Wave Thrust or Radiation Stress Note that these two expressions are of the same order in
H 2 . Therefore, the wave thrust is
16-7.1 Definitions
The concept of specific force has been defined in Section 1
Sxx = u;pgH
2[
1 + 2 sinh4nd/L
(4nd/L)
J
14-2.1 as the sum of the momentum flux through a cross
section A and the pressure force acting on this section A which gives Sxx = ?6 pgH 2 in shallow water and Sxx =
per unit weight of fluid; i.e., / 6 pgH 2 in deep water.

vz p
[-+-A
g pg
J If we consider a wave traveling over a gently sloping bed
A similar definition is introduced m the field of water
waves, the concept of" radiation stress." Radiation stress is,
by definition, the average value of the sum (Pw + p v;) with
respect to time, integrated along a vertical plane of unit
width. The velocity component perpendicular to the con-
sidered plane is V,, Pw is the pressure fluctuation due to
water waves around the hydrostatic pressure from the still
water level, i.e., Pw = p - pgz where p is the total pressure,
and p v; is the momentum flux. When the plane is vertical
and parallel to the wave crest (V = u), the radiation stress is
which is still called the "wave thrust."
16-7.2 Application to Water Waves
If one inserts the expression of Pw obtained from the
linear wave theory, the corresponding integral has an
average value equal to zero. However the second-order
term yields
1 2 4nd/L
u;pgH sinh (4nd/L)
from deep water into shallow water to the point at which
the wave height becomes equal to the deep water wave
height, as seen in Section 16-6.1 the wave thrust has to
increase by a factor of 3. Therefore, the momentum balance
requires an external force. This external force is obtained by
differences in the hydrostatic pressure. Hence, the mid water
level is set down below the still water level, as demonstrated
in the following section.
16-7.3 Wave Set-Up/Wave Set-Down;
Surf Zone Circulation
Let us consider the momentum balance in a slice of
water bounded by the free surface z = IJ, a gently sloping
bottom z = -d(x), and two vertical planes parallel to the
wave crest at x = x 0 , x = x 0 + dx. The average value of
ij = l/T J1J dt is defined and gives the variation of the mid-
water level with respect to the still water level. The total
forces exerted on the planes x = x 0 , x = z 0 + dx are the
sum of hydrostatic pressure and wave thrust. Thus,
F xo = + [Sxx + !pg(d + ij) 2 ]
d 1
Fxo+dx = - { Sxx + dx (SxJ dx + 2 pg
x [(d + ij) 2 + ~ (d + ij) 2 dx ]}
 Chapter 16: Linear Small Amplitude Wave Theories

Another external force is due to bottom pressure (since After breaking inception, the wave energy is dissipated; 233
the bottom is not horizontal). It has a horizontal component, therefore Sxx decreases and r; increases. This is the wave
set-up. The wave set-up is the elevation of the midwater
level which takes place in the surf zone. If one assumes that
all the wave energy is dissipated in the surf zone, the total
By application of the momentum theorem (Chapter 12), the force exerted on the beach (or on a wave absorber) is equal
momentum balance yields to the wave thrust in deep water, i.e.,
dSxx _)dry
Fxo+Fxo+dx+Fb~ dx +pg(d+ry dx~O

or, since r; ~ d, Therefore, it is seen that dissipating the wave energy does
dry 1 dSxx not nullify the average wave force. If the wave is reflected
dx ~- pgd dx instead of being dissipated, the wave force will be subjected
to large fluctuation but its average value, at a first order of
Since Sxx increases when the wave proceeds from deep water approximation, is zero.
to shallow water, it is seen that r; decreases, i.e., the mid water The concept of radiation stress also explains the occur-
level is set down (Fig. 16-15). rence of surf beats, i.e., the oscillations of water level on the

Figure 16-15 Effects of radiation stresses in the surf zone.

VERTICAL ELEVATION
(DISTORTED)

STILL
HORIZONTAL
DISTANCE

WAVE
THRUST

NON LINEAR MID

ELEVATION RISE
EXTERNAL PRESSURE FORCE
( HORIZONTAL COMPONENT)
 Part 3: Water Wave Theories

234 shore with a period equal to several times the wave period, 12,-------.-------.-------.-------.-----~

due to variation of wave heights within wave groups.

If the breaking waves approach the shore at an angle, the
wave thrust has a component parallel to the coast which
cannot be balanced by pressure variations. Therefore a
longshore current is generated. The piling of the water
due to the wave thrust acting towards a beach is unstable
and causes rip currents. The rip current has a momentum
flux in opposite direction to the wave thrust. Therefore, the
longshore current, instead of being parallel to the coast,
tends to meander by interaction with the movable sediments
forming the beaches. Transverse resonance effects (edge
wave) enhance and/or trigger the phenomena.

16-8 A Formulary for the Small

Amplitude Wave Theory
Figure 16-16 gives the values sinh (2nd/L), cosh (2nd/L),
and tanh (2nd/L) as functions of d/L. Tables 16-2 and 16-3
summarize a number of formulas for linear periodic waves
in deep water, in an intermediate water depth, and in
shallow water. Table 16-2 presents a set of formulas for
progressive waves, Table 16-3 for standing waves.

t VALUE
PROBLEMS
Figure 16-16 Hyperbolic functions us relative depth.
16.1 Determine the streamlines, the equipotential lines,
paths, and lines of equipressure (isobars) in a linear periodic
progressive wave.
16.4 Calculate the value of the wavelength Las a function of
the water depth din the case of a wave period T = 8 sec,
16.2 Determine the streamlines, the equipotential, lines,
10 sec, and 12 sec. Determine the minimum value of d/L
paths, and lines of equipressure in a linear periodic standing
when L can be considered as the deep-water wavelength,
wave.
and the maximum value of d/ L permitting the shallow
water approximation within 5% of accuracy.
16.3 Demonstrate byapplying the linear long wave theory
that the free surface of two-dimensional fundamental
16.5 Demonstrate that the velocity potential function for a
motion of a seiche in a basin in the form of a parabola
linear periodic progressive wave in deep water is
{d(x) = d0 [1 - (x 2 /a 2 )]} is a straight line. Demonstrate
that the free surface of the first harmonic motion is a k
= -a- e-kz cos (kt- mx)
parabola. m
Table 16-2 Periodic progressive wave, .first approximation: linear theory 235
Deep water Intermediate water depth Shallow ll"ater
d/L > 0.5 0.5 > d/L > 0.05 d/L < 0.05

C = [gL]''' = gT = f!.
gL 2nd]''' gT 2nd
Wave velocity C= [ -tanh- =-tanh- c= (gd) 1 12
2n 2n k 2n L 2n L

gT' gT 2 2nd
Wavelength L = - (Lft = 5.12 T 2 sec) L=-tanh- L = T(gd) 112
2n 2n L

k' = gm k' = gm tanh md k 2 = m 2 gd

Group velocity C 1gT

U=-=-- C [ 1 + 2md
U=-
2
---
sinh 2md
J u = c = (gd)'''
2 2 2n

Potential function Hk H k cosh m(d + z) HgT

cJ> = - --exp (mz) cos (kt- mx) cJ> = - - cos (kt - mx) cJ> = - --cos (kt - mx)
2m 2m sinhmd 2 2n

Hg H g cosh m(d + z) Hg
cJ> = - 2 k exp (mz) cos(kt- mx) </> =- - -
2 k cosh md
cos(kt- mx) = - 2 k cos (kt - mx)

Velocity components 2nH H cosh m(d + z) H

u = T 2 exp(mz) sin (kt - mx) u = - k
2 sinh md
sin (kt - mx) u = - (gjd) 112 sin (kt - mx)
2

2nH H sinh m(d + z)

w = T 2 exp(mz) cos (kt - mx) w = - k
2 sinh md
cos (kt - mx) w=O

Orbits H H cosh m(d + z0 ) HT

x - x0 = - 2 exp (mz 0 ) cos (kt - mx 0 ) X - X0 = - -
2
.
smh md
cos (kt - mx 0 ) X - X0 =-2z;;
x (gjd)''' cos (kt - mx 0 )
H H sinh m(d + z0 )
z - z0 = 2 exp (mz 0 ) sin (kt - mx 0 ) z - z0 = - . sin (kt - mx 0 ) z- z0 = 0
2 smhmd

Circles of radius: R =!!. e+m" Ellipses of: H cosh m(d + Zo) Straight lines H T ( )'''
2 horizontal amplitude: - - 1!.
Semimajor axis: 2 --s1c-.nc-h-m--:d-
2 2n d

. . . H sinh m(d + z 0 )
Sem1mmor ax1s: - - - - - -
2 sinh md

H H H
Free surface '1 = 2 sin (kt - mx) '1 = 2 sin (kt - mx) '1 = 2 sin (kt - mx)

H H cosh m(d + z) H
Pressure p = -pgz + pg-
2 exp(mz)sin(kt- mx) (1)* p = -pgz + pg sin(kt- mx) P = - pgz + pg 2 sin (kt - mx)
2 cosh md

H sinhmz0
(2)* P = - pgz 0 - pg . sm (kt - mx 0 )
2 smh md cosh md
Energy per wave
length per unit of
length of crest

Average energy flux per

unit length of crest !Pg
(
H)'gT
2 2n !pg (2H)' [ + C 1
2md
sinh 2md
J
*(1) Euler; (2) Lagrange.
 Part 3: Water Wave Theories

236 Table 16-3 Standing wave, first approximation: linear theory

Deep water Intermediate water depth Shallow water
d/L > 0.5 0.5 > d/L > 0.05 d/L < 0.05

Hk H k cosh m(d + z) Hg
Potential function = - 2 ;;; e+m' cos mx sin kt = - --
2m sinh md
cos mx sin kt =---cos mx sin kt
2 k
Hg . H g cosh m(d + z) .
= - - - e+m cos mx Sin kt =--- cosmxs~nkt
2 k 2 k cosh md

Velocity components u= - !!_ ke+mz sin mx sin kt H cosh m(d + z) H

u=--k sinmxsinkt u= - 2 (g/d) 1 12 sin mx sin kt
2 2 sinh md
H
w =- ke+m cos mx sin kt
H sinh m(d + z)
w=-k cosmxsinkt w=O
2 2 sinh md
H H cosh m(d + z0 ) . HT
Orbits x- x 0 = - e+mzo sin mx 0 cos kt
2 x - Xo = 2 sinh md Sin mx 0 cos kt x - x 0 = - - (gfd) 112 sin mx 0 cos kt
2 2n
H H sinh m(d + z 0 )
z - Zo = - 2 e+mzo cos mx 0 cos kt z - zo = - 2 sinh md cos mx 0 c~s kt z- z 0 = 0

X - Xo x- x 0 tan mx 0 H onzonta
. II"1nes _
H _
T (g/d) 112
Straight lines:--= -tan mx 0 Straight lines: - - = - -,----,.,--------,- sin mx
z- z 0 z - z0 tanh m(d + z0 ) of amplitude: 2 2n
H H H
Free surface 'f= --cosmxcoskt 'f= --cosmxcoskt 'f= --cosmxcoskt
2 2 2

Pressure p= -pgz-pg 2H e+""cosmxcoskt p= - pgz - pg 2H coshcoshm(dmd+ z) cos mx cos kt P= - pgz - pg lH cos mx cos k
ipg(~Y Lsin ipg(~Y Lsin
Energy per
wave length EK = 2 kt Ex = 2 kt EK = ipg(~Y Lsin 2 kt

E.= apg(~Y Lcos 2 kt EP = ipg(~Y Lcos 2 kt E, = ipg(~Y Lcos 2 kt

E = apgL(~)' E = ipgL(~Y E= apgL(~)'

and give the expression for the pressure p(x,z,t) and the free and demonstrate the expression:
surface. Demonstrate why the criterion d/L ::;; 0.5 is
generally considered as the limit of validity of the above gL 2rcd)
C = ( -tanh-
112

expression for . Explain why the wave refraction diagram 2rc L

must start at d/L > 0.3.
16.8 Demonstrate that the particle path in a two-dimension~
16.6 Establish the expression for p(x,z,t) for a two-di- al periodic linear standing wave is a straight line and sketch
mensional periodic linear progressive wave in intermediate the corresponding flow pattern.
water depth.
16.9 Determine the periods of free oscillation (fundamental
16.7 Consider the relationship:
and first ten harmonics) of a two-dimensional basin 10 ft
k 2 = mg tanh md long (3.048 m) and 2 ft deep (0.609 m). Determine the three

longest periods of free oscillation of a rectangular basin
10ft x 8 ft (3.048 m x 2.438 m) and 2ft deep (0.609 m).
16.10 Determine the error on the value of C corresponding
to the deep water approximation (C = gT/2n) when
d/L = 1/3. Determine the minimum value of d/L which
gives the same error for the shallow water approximation:
C = (gd) 112 Determine the common value of d/L such that
the error for the shallow water approximation equals the
error for the deep water approximation.
16.11 The pressure p(t) due to a periodic wave traveling in
intermediate water depth is recorded at a fixed location
on the sea bottom. Determine the operation which permits
one to determine the wave height from the knowledge of
this pressure fluctuation.
16.12 Establish the value of the pressure on a vertical wall
(first approximation) and determine a corresponding
method of calculation for determining the stability of a
vertical breakwater.
16.13 Consider a straight shoreline with parallel bottom
contours and a periodic wave arriving at an angle with the
deep water bottom contours. Demonstrate that at any
contour one has
sin sin
C
r:t.
= ---z;;- =
r:t. 0
constant
by application of the law of wave refraction. C is the water
wave velocity, r:t. is the angle of wave crest with the parallel
bottom contours, and subscript 0 refers to infinite depth.
16.14 It will be assumed that the principle of conservation
of transmitted energy between wave orthogonals is valid.
Moreover, it can be demonstrated that the wave breaks
over a horizontal bottom when the wave steepness
H 2nd
L-+ 0.14 tanh L
Chapter 16: Linear Small Amplitude Wave Theories
It will be assumed that this breaking criterion is still valid 237
in the case of a wave breaking at an angle on a gently
sloping beach. Now, consider a periodic wave arriving
from deep water at an angle r:t. 0 with the bottom contours,
these bottom contours being straight and parallel and
defining a very gently sloping beach. Establish a method for
calculating the angle of breaking wave crest r:t.b with the
shoreline, the depth of breaking db and the wave height of
breaking H b as a function of r:t. 0 , the deep water wave height
H 0 , and the wave period T.
16.15
1. Establish the expression for the maximum pressure
fluctuation 11p in a periodic progressive wave as a
function of depth z.
2. If a wave recorder is located at a depth z = z 0 what is
the pressure multiplication factor K to apply to the
amplitude in order to obtain the wave height.
3. Let 11p recorded be 1.33 psi (0.0933 kg/cm 2 ), T =
8 sec, z 0 = -25ft (7.62 m), d = 96ft. (29.26 m). Calcu-
late the wave height.
16.16 Establish the expression of the pressure fluctuation
due to an incident periodic wave of height Hat the toe of a
vertical wall causing total wave reflection.
16.17 In a two-dimensional periodic wave, the pressure
difference across the free surface due to capillary effect is
021]
11p = -A ox2
where 1J is the free surface elevation and x the horizontal
distance; A = 75 dynjcm for water at ooc; A = 490 dynjcm
for mercury (p mercury = 13.56 p water).
1. Establish the free surface boundary condition (linear
approximation).
2. Give the expression for the phase velocity as a
function of period and water depth.
3. Draw the curve for phase velocity as function of the
wave period for deep water.
 Part 3: Water Wave Theories

238 16.18 A wave filter is composed of wire mesh dropped into

the fluid flow. Such a filter creates negligible flow distur-
bances. However, it introduces an internal friction force F
proportional to the average velocity such that F = - KV.
The average flow motion will be considered as irrotational.
Establish the free surface condition which should be used
instead of the Cauchy-Poisson condition for the free
surface in the case where the void coefficient of the filter is
close to unity and in the case where it has a finite value .
Chapter 17
239
17-1 Introduction to the Field of
Finite Amplitude Wave Theory
17-1.1 Type of Power Series

Finite Amplitude 17-1.1.1 This chapter deals with the problem of periodic
two-dimensional finite amplitude waves, in which non-
linear effects are partly taken into account by the use of
Waves power series. The subject matter has been introduced
briefly in Chapter 15 in the general introduction to water
waves. The purpose of this chapter is to present more
advanced considerations, which form the background
for this subject matter. However, the mathematical details
are beyond the scope of this book, so that only the main
results will be presented.
It has been shown that solutions can be expressed as
power series in terms of a quantity which is small compared
to the other dimensions. The small quantity used here is
H/L for small L/d; it is the most significant parameter
in deep water. H/d is used for large L/d; it is most significant
parameter in shallow water.
In the first case (development in terms of H/L), the first
term of the power series is obtained by application of the
linear theory. These are the Stokes (or "Stokesian") wave
theories. Many series are expressed in terms of 2a/L, where
a is a parameter such that

I(2a)n
n

1
-L
_H
--
L
(a = ~ at the first order)
while others are expressed in terms of H/L directly.
In the second case, the first term of the series is already
a solution of nonlinear equations. They are called cnoidal
wave theories because they are mathematically defined
by the so-called cnoidal functions. The definition of these
functions, as well as the development of these theories,
are beyond the scope of this book. These are the essential
basic principles.

17-1.1.2 In order to illustrate this method, the following

example is presented: There exist in the American literature
 Part 3: Water Wave Theories
at least five different irrotational Stokesian wave theories
at a fifth order of approximation; all of them are based
on the same assumption, but the mathematical formulations
are different. The free surface profiles are given for three of
them (8 = mx - kt):
1J
-
H
= a Lr
n
n cos n8 (17-1)
1
m1J = rnA cos 8 + [B 22 + B 24(mAf](mA) 2 cos 28
+ [B33 + B 35 (mA) 2] (mA) 3 cos 38
+ B44(mA)4 cos 48 + B 55 (mA) 5 cos 58 (17-2)
where A is related to H/2 and the B coefficients are complex
functions of d/L.
!!__ = [1 + {3 13 (mA 0 ) 2 + {3 15 (mA 0 )4] cos 8
Ao
+ [{322 + f324(mA 0 f]mA 0 cos 28
+ [{333 + f33s(mA 0 f] (mA 0 ) 2 cos 38
+ /3 44(mA 0 ) 3 cos 48 + {3 55 (mA 0 ) 4 cos 58 (17-3)
The {3 coefficients are complex functions of d/L. This
example of the mathematical representation for steady-state
periodic waves over a constant depth illustrates the com-
plexity of a systematic comparison between wave theories
supposedly developed with the same aim. The complexity
increases with the number of possible assumptions which
can be used initially for developing a wave theory.
17-1.1.3 Consider further the case of irrotational waves.
The values of the wave characteristics depend upon the
number of terms chosen for the power series expansion,
either in terms of wave steepness H/L (Stokesian solution)
or in terms of relative height Hjd (cnoidal-type solution).
The Stokesian power series solution is not uniformly
convergent, and the validity of the solution is lost when
the relative depth djL tends to a small value (say, d/L < 0.1
for a fifth-order solution), since the coefficient functions
of d/L tend to infinity.
The same occurs in the case of the cnoidal wave solution.
There is no unique cnoidal theory; rather, the literature
contains several theories which may not be identical.
Since all cnoidal representations are truncated series, the
order of approximation is important because high-order
terms are generally significant.
There are two types of cnoidal theories. The oldest
is intuitive in nature, while the newer theories are straight-
forward and more rigorous. All are irrotational. The
primary intuitive theory is that of Korteweg and de Vries
(1895). The first and second terms of the series are deduced
but no scheme is presented for extension to higher-order
terms. The terms which are found are unique. More
rigorous theories have been developed. All are based on
a perturbation expansion. Unfortunately, even though
rigor prevails, the newer theories also diverge very rapidly.
It seems then that the high-order theory is not necessarily
better than its lower-order counterpart.
17-1.2 Vorticity and Mass Transport
17-1.2.1 The problem consists initially of solving a
problem satisfying continuity, momentum, and boundary
conditions. The motion is assumed to be periodic and the
wave profile is assumed to be that of a steady state. However,
these assumptions are not sufficient for solving the nonlinear
problem. Two more conditions are necessary. This leads
to discussion of the problem of rotationality and mass
transport (as they are related) and the arbitrariness of
water wave theories.
The arbitrariness in the calculation of wave motion
is inherent to the arbitrariness which prevails in the assump-
tions which are used in the calculation of the mass trans-
portation. The wave motion can be determined by assuming
that there is no mass transport at all. These are the closed
orbit theories, such as the exact solution of Gerstner
(1809) in deep water and the power series solutions of
Boussinesq for shallow water. As a result of this assumption,
the motion is found to be rotational and the vorticity is
in the opposite direction to the particle rotation, i.e., in
opposite direction to what should be expected physically
under the influence of a shearing stress due to wind blowing
in the wave direction (Fig. 17-1).
 Chapter 17: Finite Amplitude Waves

17-1.2.2 If one uses a power series solution in terms of

~
wave steepness H/L to the second order of approximation,
one finds that the general relationship between mass
transport U and vorticity 2ry is given by
I
2 = H2 [au _m k sinh 2m(d + z)J
2
NET MASS TRANSPORT 11 4 oz sinh 2 md
where it is recalled that H is the wave height, m = 2n/L,
L is the wave length, k = 2n/T, T is the wave period, d
is the water depth, and z is the vertical coordinate.
In the case where U(z) = 0 (Boussinesq wave)

211 (z) I = _ H 2 [m 2 k si~h ;m(d + z)J

u=o 4 smh md

NO NET MASS TRANSPORT In the Stokesian wave theories (1'/ = 0) where one assumes
that the integrating constant is such that D = 0, one has

U(z) lq=O = [ 2 si;~ md][cosh 2m(d + z) - sin:m~md]

0 In the more general case where both U and '1 are different
from zero, one can write
CLOSED ORBIT

Figure 17-1 Different kinds of assumptions on mass

ry(z) = ,u[ry(z)lu=oJ
transportation in periodic progressive waves.

The wave motion can also be assumed to be irrotational, Figure 17-2 Mass transport as a function of vorticity.
in which case a mass transport distribution is found as a WAVE DIRECTION-
result of nonlinearity. These are the Stokesian wave theories
which include Stokes (1847), Levi-Civita (1925), Struik
(1926), and Nekrassov (1951).
Even though there is a given mass transport distribution
which is a function of the vertical coordinate, the integrating
constant is often determined by assuming that the average
mass transport D is nil for the sake of continuity, i.e., a
steady flow is superimposed such that

D= Jq U(z) dz = 0
-d
where U(z) is the mass transport.
 Part 3: Water Wave Theories
242 which gives (assuming U= 0)
U(z) = (Jl + 1)[U(z)l~=oJ
where J1 is an arbitrary coefficient.
The case where J1 = - 1 corresponds to the Boussinesq
(closed-orbit) solution. J1 = 0 corresponds to the irrota-
tional theory of Stokes. If J1 > 0, the average vorticity is
in the same direction as the orbit direction, such as found
when a strong wind blows in the wave direction. The case
where J1 < - 1 gives a negative vorticity and a negative
mass transport at the free surface which can be due to a
wind blowing locally in opposite direction to the wave
propagation (a frequent nearshore occurrence). In the case
where - 1 < J1 < 0, the vorticity is in the opposite direction
and the mass transport is smaller than in the irrotational
case. These facts are schematically represented in Fig. 17-2.
Due to viscous forces, the mass transport at the bottom
is always in the wave direction. Numerous experiments
confirm this result, and the following value of the bottom
mass transport velocity can be derived:
s 2 mk
ub = u;H . h2 md
sm
17-1.2.3 Finally, it is recalled that the mathematical
representation of water wave motions are found as a
solution of a set of basic equations and assumptions.
These basic equations are the equations of continuity and
momentum, the solutions of which must satisfy given
boundary conditions. The first assumption is that of
rotationality or mass transport as they are related. However,
another condition is required. For example, for progressive
monochromatic waves, the solution sought represents
a steady state such that the potential function =
f(x - Ct), where Cis a constant equal to the wave velocity.
In this case, the solution is unique. Although the steady-
state solutions are of the same form, C is still undetermined,
and for the determination of C, another condition is
required. For example, it can be assumed that the average
horizontal velocity over a wave period at a given location
is zero, but the mass transport is then necessarily minimum.
Another condition is generally preferred. This condition
consists of assuming that the average momentum over a
wavelength is zero. In this case, another expression for C
is found which results in a different mass transport. Thus,
it is further realized that the calculation of wave theories
is subject to arbitrariness because of different assumptions
that lead to different values of C.
17-1.3 The Essential Characteristics of
Some Two-Dimensional Periodic
Progressive Irrotational Wave Theories
A brief review of some classical wave theories is now
presented.
17-1.3.1 The linear theory of Airy in Eulerian coordinates
gives the essential characteristics of the wave pattern in a
simple formulation. The free surface is sinusoidal, particle
paths are elliptic and follow a closed orbit (zero mass
transport), and lines of equipressure are also sinusoidal.
The terms in (H/L) 2 are neglected. However, nonlinear
effects such as the wave set-down can be determined from
this theory (Section 16-7).
The linear theory of Airy in Lagrangian coordinates
also gives elliptic particle paths, but the free surface and
lines of equipressure are now trochoidal (as in the deep
water wave theory of Gertsner). (A trochoid is the locus
described by a point within a circle rolling on a straight line.)
The linear long wave theory is the same as the theory of
Airy where it is assumed that d/L is small. As a consequence,
the formulas are simplified considerably. The pressure is
hydrostatic and the horizontal velocity distribution 1s
uniform. The wave velocity is simply (gd) 1i 2 .
17-1.3.2 The theory of Stokes at a second order of
approximation is characterized by the sum of two sinusoidal
components of period T and T /2, respectively. As a result,
the wave crest becomes peaked and the troughs become
flatter. The wave profile can even be characterized by the
appearance of a hump in the middle of the wave trough.
Similarly, the elliptical particle path is deformed and tends
to hump under the crest and flatten under the trough.

In this theory as in all the following wave theories, there
is mass transport as a result of irrotationality and non-
linearity. However, phase velocity, wave length and group
velocity are the same as in the linear theories. The terms
in (H/L) 3 are neglected.
The theory of Stokes at a third order of approximation
is characterized by the sum of three sinusoidal terms of
period T, T /2, and T /3, respectively. The same logical
results are found. Phase and group velocity exhibit non-
linear corrections. The coefficients of H/L which are
functions of d/L tend to infinity when d/L tends to zero
so that the theory cannot be used in very shallow water.
(The series is nonuniformly convergent.) The terms in
(H/L) 4 are neglected.
The theory of Stokes at a fifth order of approximation
is the sum of five sinusoidal terms. The coefficients of
(H/Lt are functions of d/L and tend to large values for
n > 3 even sooner than in the case of the third-order
theory, i.e., for larger values of djL. Consequently, the
fifth-order wave theory is less valid than the third-order
wave theory for small values of d/L and cannot be used when
d/L < 0.1. The terms in (H/L) 6 are neglected.
17-1.3.3 The theory of Keulegan and Patterson belongs
to the cnoidal family of water wave theories. It follows the
same physical approach as the theory of Korteweg and
de Vries. From a purely mathematical view point, there
are some inconsistencies as some third-order terms are
included while some other second-order terms are ne-
glected; however, it gives relatively good results. The
horizontal velocity component varies with depth, while
the pressure is hydrostatic.
The cnoidal wave theory of Laitone is mathematically
rigorous. At a first order of approximation, the vertical
distribution of horizontal velocity is uniform. There is no
mass transport. The terms in (H/d) 2 are neglected.
The theory ofLaitone at a second order of approximation
gives a nonuniform velocity distribution. There is mass
transport. The vertical distribution of mass transport
velocity is uniform. The second-order term becomes larger
than the first-order term as Hjd increases. (H/d is not
Chapter 17: Finite Amplitude Waves
necessarily a small parameter whereas H/L is always.) 243
The series is nonuniformly convergent. The terms in
(H/d) 3 are neglected. The results of this theory diverge
significantly from experimental results.
17-1.3.4 The solitary wave theory of Boussinesq is the
result of a purely intuitive approach. The vertical com-
ponent of velocity is initially assumed to be linearly distri-
buted from the bottom (equal to zero) to the free surface
(equal to the linearized free surface velocity 81J/8t). The
vertical distribution of horizontal velocity is assumed to
be uniform.
A correction due to path curvature (vertical acceleration)
is added to the hydrostatic pressure. The equations of
motion are linearized vertically but remain nonlinear
horizontally, i.e., convective inertia terms where the
vertical component of velocity appears are neglected,
but the product u(8uj8x) remains. The solution is then
exact. As in any solitary wave theory, 1J has always a
positive value and there is mass transport equal to the
volume of the wave above the still water level. The terms
in (H/d) 2 are neglected. The wave velocity is [g(d + H)r1 2 .
The solitary wave theory of McCowan is more rigorous
and satisfies the kinematic free surface boundary condition
exactly. It corresponds to a higher-order solution than the
theory ofBoussinesq. The vertical distribution of horizontal
velocity is nonuniform. The terms in (H/d) 3 are neglected.
17-1.3.5 The stream function theory of Reid and Dean
is also irrotational (V 2 l/J = 0) even though the use of the
stream function allows the study of rotational motion.
It is actually a numerical method requiring the use of a
computer. The solution is presented as a series of harmonic
functions with unknown coefficients determined numer-
ically in such a way that the deviation from the Bernoulli
constant at the free surface is minimum.
17-1.3.6 The formulation of God a is based on experimental
results and assumes that the particle velocity at the crest
of a limit wave is equal to the phase velocity. Accordingly
an empirical correction parameter is introduced in the
linear wave theory.
 Part 3: Water Wave Theories

244 where (a,[J) are two parameters which define the mean
position of the considered particle. The particle describes
17-1.4 The Theory of Gertsner
circles of radius r = aemfl_ The wavelength is L = gT 2 j2n.
The rotational deep water wave theory of Gerstner is The free surface and the lines of equipressure are trochoids.
exact. It is a closed orbits theory with the rotation in The limit wave is a cycloid of steepness H/L = 0.31.
opposite direction to the particle path. Its interest is Figure 17-3 illustrates some essential characteristics given
mostly historical, since it dates to 1807, although it was by this theory. The free surface corresponds to f3 = 0 and is
also developed independently by Rankine at a later date. defined by the parametric equation
It gives a simple geometric representation of the wave
motion defined by the coordinates (x,z) of a given particle, x = a - a sin (kt - rna)
such as (see Fig. 17-3; note the z coordinate system is z=-acos(kt-rna)
positive downward):
The pressure on a particle whose mean position is f3 is
x = a - aemfl sin (kt - rna) z= f3 - aemfl cos (kt - rna) independent of time and is given by the expression
2n a2
k = 2n p = pg[J + ~ k 2 e2 mfl + constant
rn=y:
T 2

Figure 17-3 Gertsner wave.

ORIGIN OF AXES
DATUM FOR

---------
(3
C Y = f3o- t'n X

~
I
DIRECTION OF WAVE /:; '(ROLLING CIRCLE
~ DIRECTION
-
OF
II:" ' RADIUS y = f3 0
ROTATION OF
GENERATING CIRCLE

z

The pressure is constant if {3 is constant, and thus all particles
at the same mean level have the same pressure as they move
about; this pressure is the same as the pressure on that
particle in its still position. The vorticity is
2m 2ka 2e 2mfJ
2ry = 1 - m2e2mfJ
Finally, the energy in a wavelength is
17-1.5 Determination of the Domain of
Validity of Finite Amplitude Wave
Theories
17-1.5.1 Knowledge of the features ofthese wave theories
is essential to the engineer interested, for example, in
calculating the forces on submerged structures such as
pilings. For this purpose, he needs reliable expressions
for the velocity field within the wave. Upon choosing a
suitable design wave (that is, mean water depth, wave
height, and wave period), he then proceeds to select a
theory to describe that wave. This choice is not easily
made, since he must evaluate the applicability of at least
a dozen theories. It may be hoped, for example, that a
theory developed to a fifth order of approximation is more
accurate than its lower-order counterparts, which may or
may not actually be the case. Unfortunately, the range of
validities of the various wave theories is not well defined.
(An attempt to classify them has been presented in Fig. 15-7.)
Also, only a few experimental results are available. Some
are presented here.
In this comparison with experiments, one can choose the
horizontal particle velocity under the crest as an important
feature, since, in applications, velocity is generally most
critical and the velocity under the crest is the greatest
attained at any depth. Furthermore, a theory which
prescribes the velocity field well is constrained to be good
for other features, such as accelerations and pressures,
a priori. Velocity profiles under the wave crest are shown
Chapter 17: Finite Amplitude Waves
in Figs. 17-4 and 17-5. A comparison of various free surface 245
profiles given by theories and experiments is also presented
in Fig. 17-6. All these comparisons are done for relatively
shallow water waves of large relative amplitude. It is
seen that the linear theory is always good at the bottom
where the boundary condition is linear and the convective
effects are relatively small. The linear theory loses its
validity toward the free surface. However, in general, the
nonlinear theories are not better than the linear theory.
Even though the numerical method of Reid and Dean
is not shown in this figure, it has been proven to be as good
as or better than the other theories when compared to
laboratory experiments. The empirical formulation of
Goda is still the best. Due to its simplicity, it appears that
an empirical relationship of this kind is the best method to
recommend from an engineering standpoint. This leads
us to discuss the selection of the "best" theory for practical
purposes.
17-1.5.2 In a narrow two-dimensional wave tank, viscous
effects in the boundary layer near the bottom induce
vorticity which diffuses upward. Also, the fact that the
wave tank is limited causes a return flow which nullifies
the total mass transport. (If the wave tank is large, transverse
effects take place, and the return flow meander.) The
theoretical approach which takes into account viscous
effects gives a mass transport which is relatively well
verified by laboratory experiments. Figure 17-7 shows
qualitatively the results of such theory.
17-1.5.3 Under storm condition, the wind stress induces
a large mass transport and vorticity in the wave motion.
This mass transport remains in a thin upper layer at the
beginning of the storm and slowly diffuses downward
under the influence of the turbulent fluctuations (mostly
due to white caps). At the same time, the mass transport
near the surface, which is in the main direction of the wave
travel at the beginning of the storm, tends to be subjected
to the Coriolis effect. Therefore the mass transport near
the free surface is at a small angle with the wave direction
and has a direction which changes with depth. Eventually
the mass transport results in a storm surge which, in
Part 3: Water Wave Theories

LONG WAVE
1st CNOIDAL

2nd CNOIDAL

,
' v~'- 2od STOKES
,+
N

~
/I
3rd STOKE' I T = 2.2 sec
H = 0.323 ft; 10.60cm
I I D = 0.619 ft; 20.31 em

I I
I I
0~----~----~~~~--~-UL---~~--~~L----L----~----~------L---~

0 0.4 0.8 1.2 1.6 2.0 2.4 2.8 3.2 3.6 4.0 4.4
u, ft /sec
0 30 60 90 120
em/sec

Figure 17-4 Particle velocity under a wave crest.

addition to being influenced by Coriolis acceleration, is
also subjected to coastal boundary effects. The storm
surge calculation method is based on the nonlinear long
wave equations (see Section 18-1.2.4) independently of water
wave effects. Therefore, it is seen that storm surge calculation
and nonlinear wave theories are two separate approaches
for what is actually a unique phenomenum resulting from
air-sea interaction in shallow water. Actually, it is a fallacy
to use irrotational or closed-orbit nonlinear wave theory
for calculating wave forces on piles, for example. The
wind-induced mass transport near the free surface could
be as high as 80% of the particle velocity given by the linear
theory. The nonlinear corrections are meaningless for an
engineer confronted with decision on what his structure
should look like. Failure of Texas towers may be attributed
to the neglect of wind-induced mass transport in water
waves.
The correct theory is a rotational wave theory with mass
transport function of wind forces, duration of wind, and
fetch (length of water surface over which the wind exerts
its action), as done in wave hindcast methods. The descrip-
tion of such a method is beyond the scope of this book.
However, a possible approach consists of finding a stream
function solution of V2 t/l = f(t/1).
 Chapter 17: Finite Amplitude Waves

1.4
y 247
/
1.2
/
/
/
4- 2nd CNOIDAL
1.0
/
McCOWAN!/

"0
......
~
0.8
/1
II
N
+
"0 1

y :
0.6

T = 1.16 sec

0.4 /1 I H =0.255ft(7.7cm)
0 =0.587ft(17.89cm)

I 1st CNOIDAL

I I
0.2
BOUSSINESQ I
I
u , ft/sec
0 30 60 90 120
em/sec

Figure 17-5 Particle velocity under a wave crest.

17-1.6 Limit Wave Steepness
Although it is beyond the purpose of this book to deeply
analyze limit wave steepness and the phenomenon of
wave breaking, the relationship between the rate of rota-
tionality and the limit wave steepness is worthwhile
mentioning.
It has been mentioned (Section 15-1.2.7) that wave-
breaking inception will occur when the wave profile
reaches a limit wave steepness H / L Imax. This limit steepness
is theoretically 0.142 for a deep water irrotational periodic
wave. Rotationality at the crest in the direction of the
wave travel such as that due to a generating wind will
reduce the limit wave steepness to a smaller value (see
Fig. 17-8). A deep sea wave steepness larger than 0.10 is
rarely encountered.
Rotationality in the opposite direction will theoretically
increase the limit wave steepness. Such a case can be
observed near the coasts when the wave travels in the
opposite direction to a wind blowing offshore. At the limit,
according to the closed-orbit Gertsner theory, the maximum
limit steepness is 0.31, but the rotationality at the crest
Part 3: Water Wave Theories

03.---------------.--------------,,--------------,--------------~--,
10

- EXPERIMENTAL L =6.4611 - 196 . 90em

-- - 3rd STOKES L = 7.16 It - 218 . 24 em
- - 5th STOKES L: 8.12 It - 247. 50 em
- McCOWAN L = oo CXl
- - I sf L AGRANGIAN L =6.63 It - 202 .08 em
- - -2nd STOKES L =6.63ft - 202.08em
- - K 8 P CNOIDAL L =7.301! - 222 .50 em
5 ...,____......AIRY L =6.63 ft -208 .08em
~BOUSS I NESQ L = CXl

,:::0.1
~
E
(,) z
Q
1-
Figure 17-6
0 ~
...J 0
Free surface elevation

, ... __ . . . . --
w

" -------
over one-half wavelenyth.
~~
T = 1. 6 sec
\.--'' -------
H = 0 .27 1 It ( 8 . 26 em) ... ~7 -
~ . =---
0 = 0. 586 It ( 17.07 em)
----._
~-
-s

2 3 4
DISTANCE FROM CREST , x , ft

0 30 60 90 120
em

is then infinity and in the opposite direction to the wave
travel. It is evident that this result of the Gertsner theory
has no physical significance.
It is seen how important it would be to establish a
general rotational wave theory and to relate the rota-
tionality and mass transport to the wind action and bottom
friction. The effect of viscous friction at the bottom has
already been subjected to investigation to some extent.
However, a general theory of irregular waves with an
arbitrary rotationality or mass transportation and quadratic
bottom friction, valid for any wave height, any wave period,
and any water depth, would be preferred. Then it will
remain to calibrate some parameters as functions of wind
stress, wind duration, and fetch length.

Figure 17-7
Periodic progressive wave with
mass transportation effect.

LARGE + SMALL ~
 Chapter 17: Finite Amplitude Waves

WAVE TRAVEL
WAVE TRAVEL 249
WIND
NO WIND

.!:!.j : 0.14 ~LH <

max
0.14
L max

Figure 17-8 WAVE TRAVEL GERSTNER WAVE

Rotationality and wave limit steepness.

WIND

ROTATION INFINITY AT THE CREST

JCI::omaxo.l4
HI: 0-31
Umax

17-2 Irrotational Waves: Methods of The free surface conditions are expressed in terms of
Calculation and Results their values at the still water level (z = 0). Developing
(x,y,O + IJ,t) in terms of power series in 11 gives
17-2.1 General Process of Calculation
It is assumed that the two unknowns, the velocity
,~,. A.(
'~'(x,y,IJ,t) = '~' x,y,O,t) + 11 [81/J(x,y,O,t)]
az +
potential and the free surface elevation IJ, may be trans-
formed into the following power series expressions with or
respect to a parameter a. This parameter is chosen arbi-
trarily but has the dimension of a length, usually taken to [Jz=~ = [Jz=o + 11 -8[8] 2 18 ]
+ -1J2 , 1-
2
8 2 +
Z z=O L Z z=O
be a half-wave height.
When this expression for is introduced into the free
= a1 + a2z + a 33 + + ann + surface conditions given in Section 16-1.3.2, one obtains
IJ = GIJ1 + a21J2 + a31J3 + ... + an!Jn + ... the kinematic condition
When this expression is substituted in the continuity
relationship V2 = 0, it is seen that each of these terms
n is an independent solution of the Laplace equation
V 2n = 0. Each 1/Jn also satisfies the fixed boundary con-
dition 8nf8n = 0 and the free surface condition.
 Part 3: Water Wave Theories

250 and the dynamic condition and

- ~ot [ + 11 ocp + .. ] + ~ [{~

oz 2 ox
( + 11 ocp + .. )}2
oz

+ { ~ ( + 11 ~~ + .. )}
2
] + gl1 = 0
Now the problem is completely formulated. When the
expressions for and 11 are substituted in the above free
surface conditions, calculations can be completed.

17-2.2 Method of Solutions

Taken in pairs, the above equations may be solved for
These relationships must be verified for any value of a
since a is arbitrarily taken. Grouping together the terms c/Jnlz=o and 11n when c/Jn- 1 1z=o and l1n- 1 are known, using
V 2c/Jn = 0 and (oc/Jn/oz)lz= -d = 0. In particular, the two
of common powers of a, it is found, respectively, that
equations which are linear in a give
( o1 0111) 2 (02 0112 0 2c/J 1
a oz + ot + a oz + Tt + 111 oz 2 a(o 1 + 0111) = 0 and o1
a( - Tt + g11 1 ) = 0
oz ot
_ o 1 0111 _ o 1 011 1) + a 3 ( .. ) + ... = O which leads to
ox ox oy oy
and [0Ot2 ~1 + g o1]
OZ z=O
= 0

This may be compared with the Cauchy-Poisson condition,

which was previously developed. The linear motion will
then be defined by = a 1, and 11 = a11 1.
In practice, the study of nonlinear problems requires
very long and tedious calculations. Sometimes, this may
+ a 3 ( .. ) + .. = 0 be reduced in the case of two-dimensional motion by using
This leads to the following equations which are independent the function W = + il/t where 1/1 is the stream function
of the value of a and i = ( -1) 112.

o 1 + 011 1 = 0 17-2.3 The Bernoulli Equation and the

oz ot Rayleigh Principle
oo/2 + 0112 + 111 o 2c/J1 _ o1. 0111 _ oo/1 0111 = 0 In the case of periodic progressive wave moving in the
oz ot oz 2 ox ox oy oy OX direction at a celerity C, the general solution for is
= P(z)F(x - Ct)
and

ocp = - CP(z)F'(x - Ct)

ot
 Chapter 17: Finite Amplitude Waves

also after division by C 2

a
ax =
'
P(z)F (x - Ct)
u-
( -C- C) 2
+ C2 +
w2
cz
2g1]
= constant

Hence it is seen that The constant can be taken equal to unity. It is seen that the
motion could be considered as a steady motion in a new
a = - Ca- =+Cu
- system of relative coordinates. In this system of coordinates,
at ax the origin of the OX axis moves at the wave celerity C
Introducing these relationships, it is seen that the Bernoulli (the Rayleigh principle). The free surface and the bottom
equation in the case of progressive waves takes the form are streamlines in this moving system of coordinates.

- Cu + t(u 2 + w2 ) + !!_ + gz = 0 17-2.4 Stokesian Theory at a Second

p Order of Approximation
After some transformations, this becomes: Some mathematical results for a periodic irrotational
2p progressive wave at a second order of approximation are
C2 - 2Cu + u 2 + w 2 + - + 2gz = C2 presented in Table 17-1.
p
These have been obtained by a method similar to the
C 2 being a constant, and p being also a constant at the free one briefly outlined in the previous sections according to
surface, the dynamic condition at the free surface becomes the assumptions described in Section 17-1.2.

Table 17-1 Periodic progressive wave: second order of approximation

Deep water Intermediate water depth

gL 2nd] 1 12
Wave velocity
2rr
C = [ -tanh-
L

Third order - [ 1 + (H
c-_ {gL 2rr)']}'! 2
---
2n 2 L

cosh m(d + z) . 3 (H)' cosh2m(d + z) .

Potential function q, = - -H C-------- Sill (kt - mx) - -m - C -------~Sill 2(kc - mx)
2 sinh md 8 2 sinh 4 md

Free surface ry = H cos (kt - mx)

2 + 1m (H)"
2 cos 2(kt - mx)
H
ry = --cos (kt - mx)
2
+ m (H)'
-
2
cosh md(cosh2md + 2)
---~-.---------cos 2(kt
4 Sillh 3 md
- mx)

= f!_ cos (kt - mx) + m(ii_) 2 - 1- - (1 + --~--)cos 2(kt - mx)

2 2 tanh md 2 sinh md
2

Midlevel elevation Ll
A = 1112- (-H2)2
~ = m(q)
2
t;n~ md (l + 2 sin~2 md)
Mass transport (H)'
(nonclosed orbits) m2 (1) No return flow: U = m2 (ii_) 2 C cosh 'm(d + z)
U = 2 Ce+ 2 m'
2 2 sinh 2 md
(2) Average mass
transport nil: U = m 2 (H)'
- C~----
1 ( cosh 2m(d + z)- -
sinh 2md)
-
2 2 sinh 2 md 2md
 Part 3: Water Wave Theories
17-3 Differences between Water Waves
and Unsteady Flow through Porous
Media
17-3.1 A Review of the Basic
Assumptions
Before concluding the study of the finite amplitude wave
theories, it is particularly interesting to establish a parallel
between unsteady flow through porous media and
irrotational water waves and to point out the essential
differences.
It is recalled that the study of the average flow through
a porous medium can generally be performed by neglecting
all inertial forces, local and convective (see Section 9-2.1.3).
Pressure and body forces always balance friction forces.
In the case of water waves, pressure and body forces
balance inertial forces and the friction forces are neglected.
As a consequence of this first equality between pressure,
body force, and friction forces, a "water wave effect"
for a flow through a porous medium is impossible. For
example, consider a hump of the free surface elevation
1J(x,t) and h(x,t) with a zero velocity at time t = 0 such as
shown on Fig. 17-9.
In the case of water waves, the potential energy is replaced
by kinetic energy as IJ(x,t) tends to zero and the kinetic
energy is in turn changed into potential energy. The free
surface elevation oscjllates around the still water level
(SWL) and the initial perturbation generates a water wave.
In the case of flow through porous medium, the potential
energy is at any time dissipated by the friction forces. The
free surface elevation tends slowly toward the still water
level. A case of practical interest where the two phenomena
DATUM
can be observed jointly is when a wave (tidal wave or wind
wave) oscillates along a pervious ground or a pervious
quay. The water table motion induced by the waves at
sea is damped very rapidly with distance.
These facts can, of course, be demonstrated and calcu-
lated exactly. A short parallel between the most typical
equations for these two kinds of motion is now presented.
17-3.2 Dynamic Conditions
Both problems consist in defining a potential function
satisfying the Laplace equation V2 = 0.
It is recalled (see Section 9-2.2.3) that the potential
function for a flow through porous medium is defined in
the case of two-dimensional motion by
8
u=-K-
8x
8
w=-K-
8z
where the average double bar 11, w is eliminated for the
sake of simplicity. It is recalled that under such a condition,
=_!!_+z
pg
In particular, = h at the free surface where z = h(x,t) by
definition. Thus the free surface equation is
h(x,t) = (x,z,t)iz=h
or agam
h(x,t) = [x,h(x,t),t]
Although the word "dynamic" would now be misused,
it has to be noted that this equation corresponds to the
Figure 17-9
Flow through porous medium
h(x, t)---> still water level (SWL).
t INCREASING Water waves: IJ(X, t) oscillates.
 Chapter 17: Finite Amplitude Waves

free surface dynamic condition for water waves: 253

- -o + -1
ot 2
[(o) (o)
-
ox
2
+ -
oz
2
] + grl = f(t)
z n

The basic difference is that in one case his proportional to , n

and in the other case IJ is proportional to ojot due to

local inertia.

17-3.3 Kinematic Condition

It is recalled (see Section 16-1.3) that the kinematic
condition for water waves is in two dimensions

A similar equation does exist for flow through porous

medium. However, due to the fact that the water fills only
the voids, and due to the change of definition of the potential
Figure 17-10 Notation for the free surface equations.
function, this equation has to be slightly modified as a
function of the void coefficient s and the coefficient of
permeability K as follows.
Let us consider an element dCJ of the free surface at time t or x), gives
t (see Fig. 17-10) and at timet+ dt. The volume of fluid
within ABCD is equal to the discharge through AB times
the interval of time dt. On one hand the volume of water
in ABCD is s dCJ dn, where s is the void coefficient and This can be written as
dn = (ohjot) dt cos a. On the other hand, the fluid discharge
through AB during the time dt is o = ( 1 _ o<P) oh
[u sin a dCJ + w cos a dCJ] dt ox oz ox

By inserting tan a = -(ohjox) and replacing u and w by and

- K(ojox) and - K(ojoz) respectively and then dividing o = ( 1 _ o<P) oh
by cos a dt, one obtains for continuity ot oz ot
o s oh o oh
-=---+-- One can eliminate h from the kinematic condition of the
oz K ot ox ox
previous section by substitution of the above two expres-
This is the kinematic condition for flow through a porous sions. This gives

(o) + (o)
media.
At the free surface z = h(x,t) = [x,h(x,t),t]. Differ- o =
_!__
2 2 o
entiating h with respect to t and x successively (* means K ot ox oz oz
 Part 3: Water Wave Theories

254 which can be written as Table 17-2 summarizes the main equations for the
!__ o =
K ot ox
i_ ( o) + ~ ( o) _ o
ox oz oz oz
formulation of a two-dimensional wave and flow through
porous medium.

since V2 = 0. When the vertical component of motion 17-3.4 Form of Solutions

ojoz is negligible, this becomes
!__ o = ~
K ot ox
( o)
ox
It is now possible to substantiate mathematically the
fact that a water wave effect is impossible in the case of
flow through a porous medium.
When = h is used this becomes the Dupuit approximation. In the case of water waves, 11 is proportional to ojot or
If the variations in hare small with respect to h, (h = d + IJ), of/Jjoz is proportional to o2 /ot2 In other words, when
the equation becomes
= f(x,z)eiat
s 011 o2 11
Kd ot ox 2 it is seen that of/Jjoz is proportional to a 2 , which is real.
which is the heat or diffusion equation of mathematical The periodic water wave solution does exist.
physics. In the case of flow through a porous medium, a wave

Table 17-2 Equations for two-dimensional waves and flow through porous mediums

Irrotational water waves Flow through porous medium

Definition of the potential function

o o
U= - - U= -K-
OX ox
o o
-K-
w = - oz W=
oz
p
Expression of the potential function QJ = f(x,z,t) QJ=-+z
pg
o 017 o 017 a ~: oh a ah
Kinematic condition at the free surface -= --+-- -= ---+--
oz at ax ax oz Kat ax ax

Dynamic condition at the free surface _ o + ~ [(8) 2+ (8)2] + g1]

at 2 ax az
= f(t)

Kinematic and dynamic condition gives

8
2
7Ji2 + g
o
oz = 0 (linearized) !!___
K&
a = [(a) + (a)
~
2

fu
2
] _ a
fu
or

~: ah a [ oh]
K. at= ax h ax
(Dupuit approximation)
 Chapter 17: Finite Amplitude Waves

effect is impossible since h is proportional to and a/az 255

is pro'portional to acjJjat.
Then a solution such as = f(x,z)ei" 1 for wave motion
would lead to an imaginary relationship between h and
unless CJ is imaginary, i.e., iCJ is then real, which means that
the motion is exponential in time instead of oscillatory.
Figure 17-9 illustrates these considerations physically.

PROBLEMS

17.1 The super elevation of the mid water level in front of a

vertical wall causing total wave reflection, (clapotis) at a
second order of approximation, is

Ll = 2m(~Y tan~ md [ 1 + 4 sin~ 2 md- 4 cos~ 2 md]

Let us consider the incident wave defined by wave height
H = 11ft (3.35 m), wave period T = 6 sec, water depth to the
still water level d = 26 ft (7.92 m). Calculate successively:
1. The value for fl.
2. The maximum and minimum water level elevation,
IJmax IJmin
3. The pressure fluctuation Llp at the bottom (using
the linear theory)
4. The maximum horizontal force on the wall and the
resulting location. (It will be assumed that the pressure
on the wall is linearly distributed between the free
surface and the bottom.)

17.2 Draw free surface lines of equipressure, particle paths,

streamlines according to the Gerstner wave theory in the
case where T = 12 sec and wave height 2a = 14ft (4.26 m).
Chapter 18 18-1 Basic Equations
18-1.1 The Continuity Equations
18-1.1.1 Integrating the continuity equation div V = 0
The Long Wave along a vertical from the bottom, z = - d(x,y), to the free
surface, 1J(x,y,t), gives

Theory f~ [au aw]

au + - dz=O
-+ -
-d ax ay az
One has
au au au
f-d -ay dz =
~ a~
f ~ -a dz = (d + 1J)-a (d + 1J) -a
-d X X y
ow
f -a dz =
~
wl'!.-d
-d z
and

u. and u. are the free surface horizontal velocity components,

ub and ub are the bottom components; a(d)jax and a(d)jay
are the bottom slopes in the X and Y directions which
permits us to relate the vertical components wb to the
horizontal components. Assuming that the horizontal
velocity distribution is uniform (i.e., u = ii = us = ub, and
u = ~ = u. = ub), and adding these terms gives:

a1J + a[u(d + '1)] + a[u(d + 1J)] = 0

at ax ay

When u = 0, the equation found in Section 3-3.1 IS

recognized.

18-1.1.2 In the study of tsunami wave generation, i.e.,

during earthquake, the water depth d is time-dependent
 Chapter 18: The Long Wave Theory

due to ground vertical displacement. It is then sufficient to be unidimensional. (In the long wave theory, the vertical 257
to replace 01]/0t by 8(17 + d)jot in the above expression to component is not considered as one dimensional.) Therefore,
obtain the continuity equation. (The momentum equation-
neglecting friction- remains the same.) p(au + u au + w au) = - op*
Ot ox oz Ox
18-1.1.3 In a river or an estuary with gentle variation of
Adding the zero quantity u[(oujox) + (owjoz)], replacing
width l(x) the continuity equation is (A = hi)
p* by its hydrostatic expression as seen in the previous
section, and integrating gives

p f~ - + -ou + -ouw) dz = - -a J~ pg(- z + 17) dz

(au 2

where Vis the average velocity in cross section A. Develop- -d ot ox oz ox -d

ing this equation and neglecting oljot (i.e., (oljox) ~ (dljdx)) If it is assumed that u is uniform along a vertical, then
gives
ou 2
au + (d + 1])-+uwl~d=
(d + 17)- z 2 + IJZ )
- g a- ( - - ,~
ot ox ox 2 -d

If w is eliminated and the continuity equation is used, then

The first two terms are easily recognized, and the last term dividing by (d + 17) and using algebra will give
is the correction due to a slight change of width dljdx.
au au 01]
at + u ox = - g ox
18-1.2 The Momentum Equation
18-1.2.1 The momentum equation for long waves can 18-1.2.3 In the two-dimensional case, a similar calcula-
be established directly, by application of the generalized tion gives:
Bernoulli equation with a local inertia force, or from the
Eulerian equation. This last approach is used. au au au 01]
Some approximations similar to the ones which were -+u-+v-=-g-
ot ax oy ax
done in the study of boundary layer (Section 4-5.2.4)
need to be done. It is assumed that w is small so that the ov ov ov 01]
-+u-+v-=-g-
terms owjot, w(owjoz), can be neglected. Also, the variation Ot ax oy oy
of w in the X, Y direction is very small, so that u(owjox),
v(owjoy) are also neglected. Then the momentum equation which, associated with the continuity equation forms the
along the Z axis is reduced to (8p*/8z) = 0, i.e., two-dimensional long wave equation.

p* = pg(-z + IJ)(+Pa) 18-1.2.4 Adding the Coriolis components -Ov and Ou

The pressure is hydrostatic.
18-1.2.2 Let us now consider the momentum equation
along the X axis. For simplicity, one considers the motion
along the X and Y axes, respectively, a quadratic friction
stress at the bottom, a quadratic shearing force due to wind
at the free surface, and an atmospheric pressure gradient
to the hydrostatic term, the equations of storm surge are
obtained. The local inertia term is sometimes neglected
Part 3: Water Wave Theories
and the storm surge is then considered as a succession of
quasi steady motion. Adding also the Corio lis components,
friction forces and body forces due to the sun-moon
attraction, the equations of tidal motion are obtained.
In the study of oceanic tide, the ground displacement due
to earth tide (earth deformation) may also need to be taken
into account.
18-1.2.5 For simplicity, consider the one-dimensional
motion. Since the wave motion at the bottom is far from
being negligible, the shearing force due to friction has in
practice a relatively great importance in long wave theory.
The long wave equation should be written (Fig. 18-1):
au au aYf r
-+u-=-g-----
at ax ax p(d + Y/)
In general, r will be assumed to be quadratic, i.e., r = pfu 2 .
The coefficient f is assumed to be independent of the
variation of u with respect to time. It is assumed that
f is the same as if the motion were steady. Consequently, f
can be expressed in terms of the Chezy coefficient Ch:
f = g/C~. The long wave equation now becomes (see
Section 14-1.1):
au au a11 g u Iu I
at + u ax = -g ax - c~ (d + Yf)
It has been seen that the friction coefficients such as f,
Ch, or n are used for gradually varied flow. Most often,
they are also considered valid for unsteady motion, such
as those due to flood waves or tidal motion in an estuary.
However, one must realize that the practical use of
such empirical formulation is only due to the lack of
further information on the friction factor for unsteady
motion. The application of the Chezy coefficient to unsteady
motion is valid provided the velocity distribution is
influenced by the friction up to the free surface as in the
case of a steady flow. This occurs when the expression for
the boundary layer thickness, which increases with the
wave period, becomes much larger than the actual depth h.
Still, density effects in a marine estuary have a nonnegligible
influence on the distribution of friction forces. They can
only be known by analyzing field records.
18-1.2.6 When the zero quantity g[d(d)jdx] gS where
S is the bottom slope (which can be positive or negative
depending upon the motion direction with respect to the
X-axis) is added to the long wave equation, (d + 17) is
replaced by the total depth h, and u is replaced by V, the
following results:
av av ah g VIVI
-+ V-+g-=+gS----
at ax ax - c~ h
Such an equation is known as the equation of Barre de
St. Venant, which is used in the study of flood waves or
tidal estuary in connection with the continuity equation
given in Section 18-1.1.3.
Figure 18-1
Notation for long wave theory.
 Chapter 18: The Long Wave Theory

Inserting this value in the Eulerian equation where the 259

nonlinear terms are neglected gives
18-1.3 The Insertion of Path Curvatures
Effect: Boussinesq Equation
Consider again the Eulerian equation along a vertical
axis,
When this is integrated in the vertical direction from a
dw 1 op point z to the free surface, a distance (d + Y/) from the origin,
-=----g
dt p oz the result is:
in which one assumes that dwjdt ~ owjot. The nonlinear
p(z) fd+~ z az'l
terms u(owjoz) and w(owjoz) are neglected, but the linear - = g[d + 1J- z] + ----dz
term owjot is nonnegligible. P z d + Yf ot 2
The pressure distribution is no longer hydrostatic. The i.e.,
vertical acceleration owjot, due to path curvatures, is
going to modify this simple approximation as follows: p(z) - [d - ] 021] (d + Yff - zz
On a horizontal bottom, the vertical component of velocity P - 9 + Yf z + ot 2 2(d + '7)
at the bottom wb is nil. At the free surface w is equal to the
velocity of the free surface itself d1J/dt = o1J/ot + u(o1Jjox) The first term is the hydrostatic pressure, the second term
(see Section 16-1.3.2). The nonlinear term u(o1Jj8x) will also is the correction due to vertical acceleration. Differentiating
be neglected in such a way that ws ~ 01J/Ot. Consequently with respect to X and inserting opjox in the Eulerian
w(z) increases from wb = 0 tows = 01J/Ot. In the most general equation along a horizontal axis gives, when the small
case, one can always say that w(z) is given by a power terms are neglected,
series such as:
01] n [ z ]"
w(z) = at~ An d + 1J
where A 1 = 1 (z = 0 at the bottom). The first term of this Now, averaging with respect to the vertical and neglecting
series (n = 1) is some terms due to the fact that 1J is small by comparison
with d gives
z 01]
w(z)=---
d + Yf at du 1 Jo [ OYf d3 1J (d + 1Jf - z 2 ]
dt = d + 1J d+~ g ax + 2
ot ox 2(d + Y/) dz
It shows that w is linearly distributed from the bottom to the
free surface. Moreover, if 1J is assumed small in comparison i.e., integrating and developing dujdt leads finally to
with d, neglecting the nonlinear term
2
z (01])
- (d + 1J) 2 ot
yields The first three terms are easily recognized from the nonlinear
long wave equation. The last term is an approximate
ow z 02 1] correction due to flow curvature. Such an equation is known
ot d + Yf ot 2 as the Boussinesq equation.
 Part 3: Water Wave Theories

260 and eliminating IJ yields

18-2 The Linear Long Wave Theory a2 u a2 u
af- gd axz = 0
18-2.1 Basic Assumptions
If one neglects the convective inertia and friction terms 18-2.3 Harmonic Solution-Seiche
in the momentum equation, one obtains
It can easily be verified that the solution can be that of
av aiJ a progressive wave such as
-=-g-
at ay IJ = H cos (mx - kt)

Similarly, if the nonlinear terms a(un)jax and a(viJ)/ay are

neglected in the continuity equation, one has u (g)l/2 cos (mx -
=H d kt)

with m = 2n/L, k = 2n/T, and L = T(gd) 112 , or that of a

standing wave (seiche)
1J = H cos mx cos kt
This set of equations characterizes the linearized long wave
theory which is valid provided IJ ~ d and 1J/L(L/d)3 ~ 1.
In the case of near two-dimensional motion vis neglected
u= H (dg)l/2 sin mx sin kt
and these equations can be written: It can also easily be verified that the linear long wave
au aiJ theory is the limit case of the small amplitude wave theory
-=-g- when d/L--+ 0 (see Section 16-3).
at ax
a aiJ 18-3 The Numerical Methods of
ax (Au) + l at = 0
Solution
where A is the cross section perpendicular to the velocity 18-3.1 The Versatility and Limits of
vector u and lis the width of the container at the free surface. Validity of Numerical Methods of
This system of equations has a great number of exact Calculation
solutions for various geometrical bottom topography.
18-3.1.1 The nonlinear long wave equations are mostly
18-2.2 The Linear Long Wave Equations treated by numerical methods and computer. For this
purpose, one transforms the set of differential equations
For example, in the two-dimensional case, with the (continuity and momentum) into a finite difference scheme.
water depth d as a constant, d = A/1. Differentiating the The calculation consists of proceeding step by step. The
momentum equation with respect to x and the continuity values of IJ(x,t) and u(x,t) at a given time t 1 and at a given
equation with respect to t, and eliminating u gives location x 1 are calculated from the knowledge of their
values a small interval away. For example, consider the
a21J aziJ
8f- gd ax2 = 0 simple linearized long wave equation which is demonstrated
in Section 18-2.1,
Similarly differentiating the momentum equation with au aiJ
-=-g-
respect to t and the continuity equation with respect to x at ax

Consider the points 1, 2, 3, 4 in a T-X diagram separated by
intervals L1x and L1t, respectively (see Fig. 18-2). Knowing
the value of '1 at points 1, 2, 3, the value of u at point 4
is obtained from the equation:
L1u L11J
-=-g-
M L1x
The known values are used to find u4 :
U4 - Uz '13 - '11
L1t =-g 2L1x
Then one proceeds step by step calculating successively
u and '1 for the entire diagram. The history of the wave
profile is so obtained as a function of time and distance.
It is easily seen that the great advantage of the long wave
theory treated by numerical analysis is its versatility.
Numerical methods are particularly convenient over
complex boundary conditions, whereas the search for
analytical solutions is beyond the scope of the best analysts.
For example, the long wave theory may be applied in a
river with variable cross section. Also, terms for bottom
friction, wind stress on the free surface, and gradients due
to complex pressure distribution can easily be taken into
account. The term for bottom friction is particularly
important for flood waves and tidal waves. Wind stress
and free surface pressure gradients can also be taken into
account numerically in the study of storm surges.
T T
Figure 18-2
-j~x 1-
X X
Two kinds of
4 X
finite difference schemes.
)(
2
~-------------------- X
SQUARE MESH
Chapter 18: The Long Wave Theory
18-3.1.2 The error and limit of validity of numerical
procedures is now discussed. The transformation of a
differential equation into finite difference involves a
systematic error. Indeed, it is known that by developing
a differential term into finite difference by a Taylor expan-
sion yields
oF L1F L1x , L1x"- 1
ox = L1x - 2 f (x) ... - -n-!- J<n>(x) ...
The first task of any numerical computation is to insure
that by taking oF/ox= L1FjL1x the cumulative error due to
the neglect of higher-order terms does not exceed the
desired accuracy. Such a study involves the search for
"stability criteria," which results in a relationship between
intervals (generally space L1x and time L1t). In the case of
high order derivative terms, stability criteria may not exist.
The high order derivative terms have then to be replaced
by a first-order derivative term of another variable, so the
number of unknowns increases. For example, o2 1J/ot 2 may
be replaced by oajot, a being equal to o1J/ot. Also, the choice
of the interval is conditioned by the cumulative error,
cost of the computing time, and the "round-off error."
The round-off error is due to the fact that any numerical
calculus is necessarily done with a limited number of
figures or "digits." For example, most calculation done on
computers is done with 8 digits, sometimes 16 digits if one
used "double precision," or even more. But the increasing
cost of computing time offsets this advantage.
X
X X l(
~-----------------------4--x
STAGGERED MESH
 Part 3: Water Wave Theories

262 In brief, the numerical implementation of the long wave and then differentiating so that oc 2 jo* = co(2c)jo*, and
theory always involves an inherent error, which increases dividing the second equation by c yields
with time andjor distance.
au au a2c
The study of the propagation of a bore, or an undulated at + u ax + c -a; = - gS
wave, or the wave created by the breaking of a dam over a
long distance is unreliable even if the stability criteria is a2c au o2c
satisfied because of the cumulative effect of the error. -+c-+u-=0
at ax ax
However, the study of a tidal wave in an estuary, or even
of a flood wave with gentle variation of depth, is possible. Adding and subtracting these two equations gives
Similarly, the propagation of a breaking wave (bore) over a
ot (u 2c)
a
+ (u c) ox (u 2c) =
a steep beach, i.e., over a short distance because of the -gS
steep slope, may give reliable results.
It is now recalled that the total derivative of an expression
18-3.2 The Method of Characteristics A(x,t) with respect to time is
18-3.2.1 The characteristic equations are now established. dA(x,t)
oA dx oA
Consider the long wave equations in the following form: ~=at+ OX dt
au au ary Thus, the left-hand side of the above equation is the
Momentum: - + u- = -g-
at ax ax total derivative

ary + a[u(d + ry)] d

Continuity:
= 0 -dt (u +
-
2c)
at ax

Then consider the zero quantity: provided that

dx
-=
dt
u +
-
c

This means that along a line of slope dxjdt = u c, the

When this is added to the right side of the momentum relationship
equation, the result is
d
au au a[g(d + n)] dt(u2c)=-gS
- + u- + = -gS
at ax ax
will apply (see Fig.. 18-3).
Since o(d)jat = 0, it can be added to the continuity equation. The lines of slope dxjdt = u + care called the advancing
After the continuity equation is multiplied by the constant g or positive characteristics. The lines of slope dxjdt = u - c
the result is are called the receding or negative characteristics.
It is seen that from a given point (x,t) in the diagram,
o[g(d + IJ)] + a[ug(d + I])] = 0 two lines of slopes u + c and u - c can be drawn.
at ax
Define c = [g(d + ry)] 1 12 which has the dimension of a 18-3.2.2 An example, the "dam-break problem," is now
velocity. Inserting c2 = g(d + ry), into the above equations presented: In the case of a horizontal bottom S = 0 and
 Chapter 18: The Long Wave Theory

u = 0 is dxjdt = 0 + c, i.e.,
(0 + 2c)l 1 =o = (0 + 2c)lt=t
and

~I c = dx = (gd)1 f2
dt
'--------x '--------.-x
18-3.2.3 In general, the variation of u and 1J (or c) are
Figure 18-3 Advancing and receding gentle enough in order that a finite interval method can be
characteristics. applied. The basic characteristic equations are then written
for finite intervals .1x, 11t:
(d/dt)(u 2c) = Oi.e.,u 2c = constantvaluesK 1 andK 2
.1x
along lines of slope dx/dt = u c. Consider the case .1(u 2c) = -gS .1t along ,1( =uc
of a vertical wall of water which is suddenly released as in
the case of a dam breaking (see Fig. 18-4). It is seen that when The time history of the wave evolution can then be
t = 0, u = 0, and c = (gd) 1 ' 2 , where dis the water depth and determined step by step as follows. The values of u(x,t 1 )
the constants K 1 and K 2 are and 1J(X,t 1 ) are given for a wave at a timet = t 1 (see Fig. 18-5).
u + 2c = K 1 = 2(gd) 112 The values of u and c are calculated at regular intervals
u- 2c = K 2 = -2(gd) 112 .1x. The characteristic line of slope .1xj 11t = u 1 + c 1 is
drawn from point 1 and the characteristic line of slope
At time t = t 0 , the water wall collapses. At the down- .1x/M = u 2 - c 2 is drawn from point 2.
stream water tip: 1J = - d and c = 0. Hence the speed of the Their intersection at point 3 defines x 3 , t 3 graphically.
water tip is dx jdt = u( + 0) and Then, by applying the characteristic equation along these
(u + 2c)lr=o = (u + 2c)lr=r lines, the values ofu 3 and c 3 (and consequently 1} 3 ) are found
from the two equations
becomes 0 + 2(gd) 1 ' 2 = u.
The water tip travels at a speed 2(gd) 1i 2 . The speed of the u3 + 2c 3 = u1 + 2c 1 - gS 1(t 3 - t 1 )
rarefaction wave traveling upstream where 1J = 0 and u3 - 2c 3 = u2 - 2c 2 - gS 2 (b - t2)

Figure 18-4
Notation for the dam break problem
AB = 2CD.
Part 3: Water Wave Theories
Figure 18-5 Application of the method of
characteristics.
In the particular case of the Fig. 18-5, t 2 = t 1 . Similarly, u 5
and c 5 are found from the points 2 and 4, u6 and c6 from
the points 3 and 5, and so on.
18-3.2.4 By this process of calculation, it is seen that the
state of a wave at a given point 1 (xd 1) (see Fig. 18-6)
has an influence upon the state of the wave at any other
point between the characteristic lines issued from that
point. Such characteristic lines define the domain of
influence of this point. Similarly, the state of the wave at a
given point 3 depends solely upon the state of the wave
under the two characteristic lines crossing at point 3.
Such a domain is the domain of dependence of point 3.
Any disturbance arising over a finite distance (x 1 - x 2 )
will have an influence on the water behavior within a
domain defined by a negative characteristic issued from
x 1 and a positive characteristic from x 2 .
18-3.2.5 In the case of tidal estuary, the bottom friction
and variation of cross section must be taken into account.
The momentum equation can easily be modified for taking
into account the bottom friction as in Section 18-1.2.6. The
continuity equation can also be modified for the variation
of cross section as in Section 18-1.1.3. The same calculations
done previously lead to
~(u 2c) = -[gs + ~~ (~Y u~ :~] ~t
m which case the method of characteristics will apply
similarly.
18-3.3 Tidal Bore
18-3.3.1 When two positive characteristics dxjdt = u + c
cross each other (see Fig. 18-7), one set of values can be
found from points a and b: u 2 and c2 ; another set of values
can be found from points a and c: u 1 and c 1 .
The two values for u and the two values for c at the same
location indicate a discontinuity or a vertical wall of water.
There is bore inception.
18-3.3.2 The line dxjdt = W, where W is the speed of the
bore, is a line of discontinuity for the mesh formed by the
positive and negative characteristics. At the locus of the
bore there are five unknowns to be determined, namely:
u 1 , 11 1 on the low side of the bore, u 2 , 1J 2 on the high side of
the bore, and the bore velocity itself, W. Since W > u 1 + c 1 ,
u 1 and 1] 1 are determined by application of the method of
Figure 18-6
Domain of influence
and domain of dependence.
 Chapter 18: The Long Wave Theory

3. Since u 2 + c2 always exceeds W, the positive char- 265

acteristic equation which merges with the bore line
at the considered point (x,t) is [c 2 = (gh 2 ) 112 ]:
u2 + 2c 2 = u0 + 2c 0 - gS(t 2 - t0)

The wave profile on the high side of the bore is deter-

mined afterward by making use of u2 and c2 by applying
the method of characteristics in a straightforward manner.
b The problem of a limit solitary wave (see Section 18-4.2)
travelling on an horizontal bottom and reaching a 1/10
~---------------------------4-- bottom slope is treated on Fig. 18-9, as an example of
applicability of the method of characteristics. The successive
Figure 18-7 Bore inception.
wave profiles are presented on Fig. 18-10 and are obtained
by interpolating between the values of IJ at the crossing
characteristics directly from the points a and c (see Fig. 18-8). of characteristic lines by lines defined by t = constant.
u2 , 1] 2 and W are determined from the three following
equations: 18-3.4 The Direct Approach to
1. The momentum equation for a moving hydraulic Numerical Solutions
jump (see Section 12-4.6) which after some transfor- The application of the method of characteristics requires
mations gives: the solution of four unknowns at each point of crossing
ipg(h~ - hi)= ph 1 (u 2 - u 1 )(W- u 1 ) lines, namely, x, t, u, and '7 In the case where x and t are
specified a priori, only two unknowns remain: u and IJ.
2. The continuity equation for a moving hydraulic For this purpose the long wave equations can be treated
jump: directly by a finite difference process. The intervals llx
and Llt, i.e., the locations of x and t, are specified inde-
pendently from the characteristic lines. For this purpose,
a square mesh or a staggered mesh can be used (see Fig. 18-2).
Figure 18-8 Bore propagation.
18-3.4.1 In the case of the square mesh method, u and IJ
at point 4 can be deduced from the value of u and '1 at
points 1 and 3 directly as follows (see Fig. 18-11). From
the continuity equation one has {(2) = t[(1) + (3)]}:

M
'74 = IJz + 2 Llx [u(d + '7)1 3 - u(d + '7)1 1]
and from the momentum equation one has:
 Part 3: Water Wave Theories
266
Figure 18-9 Application of the method of characteristics to
a wave breaking on a slope.
The calculation will proceed by calculating u 5 and IJs
from points 2, 3, 6, and so on.
18-3.5 The Stability Criterion
The error made in transforming the differential equation
into a finite difference equation may cause a cumulative
error which may blow up as t = L M increases. The finite
difference method is stable provided the point (x,t) 4nder
consideration is within the domain of dependence formed
by the intersection of the characteristics lines from the
points which are used for its determination.
Consequently, since in general (u + c) 1 > (u - ch (see
Fig. 18-11), the criterion for stability is /).t < (f).xju +c).
6 7 8 9 10
10
, ,woz~
The variation of u and c with respect to x and t also has
to be gentle enough in order that the finite difference terms
have a value close to the differential.
18-4 On Some Exact Steady-State
Solutions
In general, an unsteady-state solution is a typical
characteristic of the long wave theories, i.e., the wave
profile changes its shape as the wave proceeds. This fact
is inherent in the long wave equations. However, the
insertion of bottom friction and a vertical acceleration term,
even approximated, into the long wave equations permits
 Chapter 18: The Long Wave Theory

l~
10

II II I I L! I I I v

I
6
I

LUJ 11-~ 15 T=
4
.
5 I I

J_LLJ},;. 11/<~ I
~ I I' SPILLING BREAKER ILINI

~~ i

10 -9 -8
~I
-7 -6 5 -4 -3 -2 0
I
4 5 6 7 9
X

Figure 18-10 Time-history of the wave profiles breaking

on a "one-tenth" bottom slope.

the finding of some special steady state solutions. They are Figure 18-11 Notation for application of the
the "monoclinal" wave, the solitary wave, and the cnoidal square mesh method.
wave. Only the first two cases are analyzed in this book.
It is recalled that the assumption of a steady state solution
results in the search of a function such as

:} = f(x- Ct)

where C is a constant. Consequently, 6
a
-=-C-
a
ot ox
 Part 3: Water Wave Theories

268 depends upon whether the water depth is larger or smaller

18-4.1 Monoclinal Waves
Consider the long wave equation with a bottom friction
term such as r = pfu 2 (see Section 18-1.2.6):
than the critical depth, as shown by Fig. 18-12. These
waves are called monoclinal waves or uniformly progressive
flow. This theory is particularly suitable for the study of
flood waves in rivers.

l
oh + o(hu) = 0
at ax 18-4.2 Solitary Wave Theory

ou + u ou + g oh = gS _ jf_ u Iu I
Consider the long wave equation for a horizontal bottom
ot ax ox c~ h with a vertical acceleration term as it has been established
in Section 18-1.3:
Inserting the relationship o/8t = - C(ojox) in the continuity
equation gives (u - C)(8h/8x) + h(oujox) = 0 which inte- ou au 01] d + 1] 03 1]
grates to (u - C)h = A(constant). The momentum equation at + u ax + g ox+ -3- ot 2 ox = 0
becomes
01] + o[u(d + 1'/)] = 0
(u _ C) au
ox
+ goh =
ox
g(s _ ~)
C~h
at ax

Inserting the relationship ojot = - C(()jox) gives

Eliminating oujox between these two equations gives:

[g - (u -h qz]:~ = {s- ~;h]

By using the value u - C = A/hand then solving for ohjox, and
it is found that: a
ox [-Cry+ (d + ry)u] = 0
oh S- [(Ch + A) 2 /C~h 3 ]
ox 1 - A 2 jgh 3
The quantities between brackets are independent of x
This differential equation has a number of solutions, how- and consequently equal to constants. It is seen that these
ever, some are without physical significance. The solution constants are nil since u and 17 tend to zero when x ----> oo.

Figure 18-12
M onoclinal waves.
 Chapter 18: The Long Wave Theory

From the second equation one obtains u = Ct]/d + f/.

When this is substituted into the first equation, the results
are d d

Figure 18-13 Solitary wave.

By considering tJ small with respect to d and developing
the square root as (1 + a) 112 ~ 1 + !a, the above equation
can be solved for C such that The equation can still be written:
t] d2 02t))]1/2
c = [ g(d + 'Y/) ( 1 + 2d + 3t] ox 2 ~: = (:3)1/2tJ(H- '1)112

When the free surface curvature is negligible, the value for C After separation of the variables, it can be integrated in the
becomes form

c ~ (gd) 112 ( 1 + ~ ~) fo~ IJ(H dt] fx ( 3 )1/2

- '1)1 /2 = o d3 dx
which can also be obtained directly by application of the
which gives for the wave profile :
momentum theorem.
In this case, since one has assumed C to be a constant H
and since the wave profile remains unchanged with time, 'Y/ = cosh 2 [(3H/d) 1i 2 (x/2d)J
one must have:
as shown in Fig. 18-13.
3t] d 2 o2tJ H
2d + 3t] ox2 = constant = d

H is a constant, which is specified in the following. This PROBLEMS

equation can be integrated as follows:
18.1 Demonstrate that the two-dimensional linear long
02t] 3t]
OX2 = 2d3 [2H - 3t]] wave equation over a horizontal bottom is

Since OtJ/ox = 0 when tJ -+ 0,

otJ o2tJ 3tJ
2 ox ox2 dx = d3 (2H - 3tJ) dt] 18.2 Demonstrate that the two-dimensional linear long
or wave equation over a slow varying complex bottom
topography for periodic motion is
[ df/]2 = 3f/2
dx d3
(H - 'Y/)
i_
ox
[d o] + i_ [d o] = - k2
ox oy oy g
<P
It is seen that tJ = H gives dt]/dx = 0 and corresponds to
the top of the wave; consequently, H is the wave height. where k is a wave number and d = d(x,y) is the water depth.
Part 3: Water Wave Theories

18.3 Consider a square mesh of mesh size A in a X, Y tion of two-dimensional long waves into a finite difference
system of coordinates. Demonstrate that the linear long system corresponding to intervals Ax, At forming a square
wave equation for periodic motion for a finite difference mesh. Do the same in the case of a staggered mesh.
scheme follows the relationship (m = 2n/L):
18.9 Demonstrate that the equation of Barre de St. Venant
-4</Ji,j + </Ji,j+ 1 + <Pi+ 1,j + </J;,j-1 can be written:
(Am) 2
+A. gd 'f'l,J-=0
1 +--A..
'f/1- ,]

18.4 Let us consider the linear long wave motion defined by where
the free surface equation, such as: m = 2n/L, k = 2n/T
a= [ Q2 _ g~ 3 ]
ry(x,y,t) = cos kt cos mx + sin kt cos my
This defines a motion obtained by the superposition of f3 = QA
two standing waves at right angles. y = A3
1. Establish the equation of the lines of constant wave
amplitude. F = A3 [~!!!_- S- IQJQIJ
g [2 dx C~A
3
2. Determine the points of maximum amplitude.
3. Is there any amphidromic point (point of zero Give the definition for .
amplitude)?
Describe the motion around it.
4. Draw a sketch of the curve of constant amplitude REFERENCES FOR PART THREE
within a square defined by 0 < x/y < L/2.
Basset, A. B., A Treatise on Hydrodynamics, Vols. I and II. Re-
printed by Dover Publications, Inc., New York, 1961.
18.5 Establish the equation of motion for periodic linear
long wave in a corner defined by a depth d = est, and width Beach Erosion Board, A Summary of the Theory of Oscillatory
Waves. Technical Report No. 2, U.S. Government Printing
b = b0 x. Office, 1942.
Biesel, F., Equations generales au second ordre de Ia houle irreguliere.
18.6 Consider the energy diagram as shown on Fig. 14-5. La Houille Blanche, May 1952.
Indicate the modification which needs to be made to such Birkhoff, G., Hydrodynamics. Princeton University Press, 1950.
a figure in the case of unsteady motion. Bouasse, H., Houles, rides, seiches et man!es. Librairie Delagrave,
Paris, 1924, pp. 92-145.
18.7 Demonstrate that the long wave equations can be Boussinesq, J., Theorie de !'intumescence liquide appelee onde
transformed as u + 2c = constant and u - 2c = constant solitaire ou de translation se propagant dans un canal rectangu-
along a line of slope laire. Institut de France, Academie des Sciences, Comptes
Rendus, June 19, 1871, p. 755.
dx gS
- =u+c+ --:-c~-:---~ Boussinesq, J., Essai sur Ia theorie des eaux courantes. Institut de
dt - - (ojox)(u 2c) France, Academie des Sciences, Memoires presentes par divers
savants, 23, 1877.
18.8 Transform the differential equations which are used Broer, L. J. F., On the propagation of energy in linear conservative
in the method of characteristics for the study of the propaga- waves. Appl. Sci. Res., A2: 447-468, 1951

Carr, J. H., and Stelzriede, M. E., Diffraction of Water Waves by
Breakwaters in Gravity Waves. U. S. National Bureau of
Standards, Circular 521, pp. 109-125.
Chappelear, J. E., Shallow water waves. J. ofGeophys. Res. 67, 1962.
Coulson, C. A., Waves, 4th ed. Oliver and Boyd, Edinburgh, 1947.
Dane!, Pierre, On the Limiting Clapotis. U. S. National Bureau of
Standards, Gravity Waves, NBS Circular 521, 1952, pp. 35-38.
Dean, R. G., Stream Function Wave Theory-Validity and Applica-
tion. Specialty Conference on Coastal Engineering, ASCE,
1965.
Druet, C., Nomographic chart for determination of the monochro-
matic wave type in the region of foundation of a designed
hydrotechnical structure. Paper S, 11-I, 21st International
Navigation Congress, Stockholm, 1965, pp. 183-201.
Dubreuil-Jacotin, J. L., Sur les ondes de type permanent dans les
liquides heterogenes. Rendiconti della Accademia Nazionale dei
Lincei, Ser. 6,15: 814--819, 1932.
Dubreuil-Jacotin, J. L., Sur Ia determination rigoureuse des ondes
permanentes periodiques d'ampleur finie, Journal de Mathemati-
ques Pures et Appliqw!es, Ser 9, 13: 217-291, 1934.
Eckart, C., The Propagation of Gravity Waves from Deep to
Shallow Water. U. S. National Bureau of Standards, Gravity
Waves, NBS Circular 521, 1952, pp. 165-173.
Freeman, J. C., and Le Mehaute, B., Wave breakers on a beach
and surges on a dry bed. Proc. Amer. Soc. Civ. Eng., 90: 87-216,
1964.
Friedrichs, K. 0., On the derivation of the shallow water theory,.
Appendix to the formation of breakers and bores by J. J.
Stoker. Comm. Pure Appl. Math. I: 81-85, 1948.
Gerstner, F., Theorie der Wellen, Annalen der Physik, 32, 1809.
Goda, Y., Wave Forces on a Vertical Circular Cylinder. Report
No. 8, Port and Harbor Technical Research Inst., Japan, 1964.
Ippen, A. T. (editor), Estuary and Coastline Hydrodynamics.
McGraw-Hill, New York, 1966.
Johnson, J. W., O'Brien, M. P., and Isaacs, J. D., Graphical
Construction of Wave Refraction Diagrams, Hydrographic
Office, Navy Department, Publication No. 605, 1948.
Keller, J. B., The solitary wave and periodic waves in shallow water.
Comm. Appl. Math., December 1948.
Keulegan, G. H., and Patterson, G. W., Mathematical theory of
Irrotational translation waves. U.S. NBS J. Res. 24, 1940.
Kinsman, B., Wind Waves. Prentice-Hall, Englewood Cliffs, New
Jersey, 1965.
Chapter 18: The Long Wave Theory
Koh, R. C. Y., and Le Mehaute, B., Wave shoaling. J. Geophys.
Res., 71: 2005-2012, 1966.
Kotschin, N. J., Kibei, I. A., et a!., Theoretische Hydrodynamik.
Adad. Verlag, Berlin, 1964.
Korteweg, D. J., and de Vries, G., On the change of form of long
waves advancing in a rectangular canal on a new type of long
stationary waves. London, Dublin and Edinburgh, Philosophical
Magazine, Ser. 5, 39: 422, 1895.
Kravtchenko, J., and Daubert, A., La Houle a trajectoires fermees
en profondeur finie. La Houille Blanche, 12, 1957.
Laitone, E. V., The second approximation to cnoidal and solitary
waves, J. Fluid Mech., 9, 1960.
Lamb, H., Hydrodynamics. Dover Publications, New York, 1945;
Cambridge University Press, 1932.
Le Mehaute, B., Mass transport in cnoidal waves. J. Geophys. Res.,
73: 5973-5979, 1968.
Le Mehaute, B., Divoky, D., and Lin. A., Shallow water waves: A
comparison of theories and experiments. Proc. Amer. Soc.
Civ. Eng. (lith Conf. Coastal Eng.) 1, 86--107, 1968.
Le Mehaute, B., and Webb, L., Periodic gravity waves over a gentle
slope at a third order of approximation. Proc. Amer. Soc. Civ.
Eng. (8th Cong. Coastal Eng.) 1: 23-40, 1964.
Le Mehaute, B., Theory of Explosion-Generated Water Waves.
Advances in Hydroscience, Vol. 7, Academic Press, New
York, 1971.
Le Mehaute, B., Theory of wave agitation in a harbor. Trans.
Amer. Soc. Civ. Eng., 127: 369-383, 1962.
Longuet-Higgins, M. S., and Stewart, R. W., Radiation stress and
mass transport in gravity waves. J. Fluid Mech., 13: 481-509,
1962.
Longuet-Higgins, M. S., Mass transport in water waves. Phil.
Trans. Roy. Soc., A 245: 535-581, 1953.
Mei, C. C., and Le Mehaute, B., Notes on the equations of long
waves over an uneven bottom. J. Geophys. Res., 71: 393-400,
1966.
McGowan, J., On the solitary wave. London, Dublin and Edinburgh,
Philosophical Magazine, Ser. 5, 32: 45, 1891; On the highest
wave of permanent type. Ser. 5, 38: 351, 1894.
McNown, J. S., Waves and Seiches in Idealized Ports. U.S. National
Bureau of Standards, Gravity Waves, NBS. Circular 521, 1952.
Miche, A., Mouvements ondulatoires de Ia mer en profondeur
constante ou decroissante. Annales des ponts et chaussees,
1944, pp. 25-78, 131-164, 270-292, 369-406.
 Part 3: Water Wave Theories

272 Mitchell, J. H., The wave resistance of a ship. London, Dublin and
Edinburgh, Philosophical Magazine, Ser. 5, 45, 1898.
Milne-Thompson, L. M., Theoretical Hydrodynamics. Macmillan,
London, 1938.
Munk, W. H., The solitary wave theory and its application to surf
problems. Annals New York Acad. Sci., 51: 376-424, 1949.
Munk, W. H., and Arthur, R. S., Wave Intensity Aloi).g a Refracted
Ray. U. S. National Bureau of Standards, Gravity Waves,
NBS. Circular 521, 1952.
Pierson, W. J., Jr., The Interpretation of Crossed Orthogonals in
Wave Refraction Phenomena. Technical Memorandum No. 21,
Beach Erosion Board, Corps of Engineers.
Putnam, J. A., and Arthur, R. S., Diffraction of water waves by
breakwaters. Trans. Am. Geophys. Union, 29: 1948.
Rouse, H., Fluid Mechanics for Hydraulic Engineers. McGraw-
Hill, New York, 1938.
Stokes, G. G., On the theory of oscillatory waves. Trans. Cambridge
Philosoph. Soc., 8, 1847; and Suppl., Scientific Papers, 1.
Struik, D. J., Determination rigoureuse des ondes irrotationnelles
periodiques dans un canal a profondeur finie. Mathematische
Annalen, 95: 595--634, 1926.
Suquet, F., Remarks on graphical computation of wave refraction.
International Association for Hydraulic Research, Grenoble,
1949.
Sverdrup, H. U., A Study of Progressive Oscillatory Waves in
Water; A Summary of the Theory of Water Waves. Technical
Reports No. 1, 2, Beach Erosion Board, Office of Chief Engine-
ering, U. S. War Department.
Thomas, H. A., The Propagation of Stable Wave Configurations in
Steep Channels. Carnegie Institute ofTechnology, Pittsburgh, Pa.
Thomas, H. A., Hydraulics of Flood Movements in Rivers. Carnegie
Institute of Technology, Pittsburgh, Pa., 1937.
Thomas, H. A., Propagation of waves in steep prismatic conduits.
Proceedings of Hydraulics Conference, University. of Iowa,
Studies in Engineering, Bulletin 20, 1940.
Ursell, F., The long-wave paradox in the theory of gravity waves.
Proc. Cambridge Philosoph. Soc., 49, 1953.
VanDorn, W. G., Tsunami. Advan. Hydrosci., 3: 1--48, 1965.
Wehausen, J. V., and Laitone, E. V., Surface waves, In Handbuch
der Physik, Springer-Verlag, Berlin, 1960.
Wiegel, R. L., Oceanographical Engineering. Prentice-Hall, 1964.
Whitham, G. B., Mass momentum and energy flux in water waves.
J. Fluid Mech., 12: 135-147, 1962.
 Appendix A A-1 Introduction to a Real Sea State
273

A-1.1 A New Approach to Water Waves

A brief introduction to the hydrodynamics and mathe-
matics of periodic and other analytically defined waves
Wave Motion in a heavy fluid bounded by a free surface has been presented
in Part Three. The developments along these lines have
as a Random Process been continued for well over a century.
Any observation of a real sea reveals a continuously
changing random pattern of bumps and hollows. Waves
of different lengths travel at different speeds combining
and recombining in constantly changing patterns even
if they are all unidirectional. When multidirectional waves
are present the patterns created are even more complex.
It was thought for a long time, even by such men as Rayleigh
and Stokes, that this apparently chaotic process was
beyond adequate mathematical description. The best that
could be done appeared to be the choice of a mean wave
height and a mean wavelength, followed by the application
of the classical wave theories. To the mathematician the
classical wave theories themselves were fascinating and the
apparent difficulties of dealing with a real sea did not
concern him.
In comparatively recent times an approach to the
understanding of real seas has been developed. This
approach is based on the combination of statistics, Fourier
analysis, and hydrodynamics. The theories of statistics
have to be used to determine stable parameters for de-
scribing a random sea state; then Fourier analysis is used
to break down the random process into harmonic com-
ponents whose behavior can be analyzed by using classical
hydrodynamic theories of wave motions.
Recent progress owes a great deal to the development of
research on the statistical analysis of random noise by
communications engineers. The study of sea waves has
developed as a combination of time series analysis and
statistical geometry governed by the laws of hydrodynamics.
This section presents an introduction to this nondeter-
ministic aspect of waves. Probability concepts and spectrum
theory will be discussed.
 Appendix A: Wave Motion as a Random Process
274
A-1.2 Some Definitions and Stability
Parameters
A part of a typical wave record is shown in Fig. A-1.
The concepts of surface ordinates, "wave height," "wave
periods," maxima and minima, and envelopes are illustrated.
Some of the statistics of the various parameters which can
be obtained from such a wave record as Fig. A-1 will be
discussed. Clearly, if the wave record is very long it is
impractical to keep it in its original form. Methods of
condensing the gross details of the wave record are required
whereas inevitably, after such a process, much detail will
be lost.
The condensations of the real sea state need to have the
property of stability. The term stability is used to describe
a characteristic which does not change too much if the
observation is repeated. For example, suppose two wave
recorders were placed in the open sea at a distance of, say,
200 ft apart. The water surface-time history of the two
records taken at the same time would be completely
different. The sea surface records themselves are unstable.
On the other hand, such things as the average wave height,
Figure A-1 A typical wave record at sea.
~"'""ATE,
WAVE PERIOD /'
mean square water surface fluctuation, etc. of the two
records would be very close to being equal (so long as the
records were of reasonable length). Such statistical prop-
erties are said to be stable.
Probability densities and distributions of sea surface
parameters, together with the spectral (or variance-
frequency) distribution of the sea surface, have been found
to be concise and useful properties of this process. The
spectrum is a form of probability distribution and has
very desirable stability characteristics. The spectrum retains
much information on wave amplitudes and "periods"
but loses all information on phase position. Probability
distributions, on the other hand, lose all information on
wave periods if "wave height" probabilities are computed
or VIce versa.
The description of sea states in terms of spectra has
already been used extensively by naval architects to predict
ship performance. Spectral and other probability concepts
are used by mechanical and aeronautical engineers to
handle vibration problems. Many branches of geophysics,
such as turbulence, seismology, and tide analysis, use the
basic communications theory to describe the processes
and analyze data.
 Appendix A: Wave Motion as a Random Process

A summary of the most important probability distri- AMPLITUO 275

butions and an exposition of spectral representation of A3

i
waves as random processes is now presented.

AI A4
A-2 Harmonic Analysis A2
I

A-2.1 The Concept of Representation of I l I r As

I
Sinusoidal Waves in the Frequency
Domain FREQUENCY
Ill 112 k3 k4 k5
Consider a sinusoidal wave profile given by
Figure A-3 Amplitude spectrum of a periodic
Yf = A cos (mx + kt + ~>) wave of arbitrary shape.

where A is the amplitude, m = 2n/L and k = 2n/T, where

L and Tare wavelength and period, respectively, and B is a
phase angle. If this wave is observed at a point as a function
of time and the origin is chosen such that the initial phase
position is zero, it can be written Yf = A cos kt. This wave
can be described as a wave of amplitude A and frequency k
(in radians). It can be represented as in Fig. A-2, which is
referred to as an amplitude spectrum.
Suppose a wave of arbitrary shape, but having a period
T (frequency 2n/T) is to be represented on an amplitude
spectrum such as Fig. A-2. Recall that almost any periodic
function can be represented as a Fourier series of funda-
Figure A-2 Amplitude spectrum of a
sinusoidal wave.
AMPLITUDE
mental period T and harmonics, so that
Yf = L An COS (nkt + ~>n)
This wave would have an amplitude frequency spectrum
appearing as in Fig. A-3.
If a water surface is described by a combination of a
number of waves having different periods (frequencies)
it could be represented as a diagram such as Fig. A-3,
having amplitude and frequency components at many
different places.
A-2.2 Review of Fourier Analysis
A-2.2.1 The immediate temptation is to take a wave
record such as Fig. A-4, assume that it can be represented in
the interval 0 < t < T1 by a Fourier series:

T! -i + L -r;- -r;-
A [ 2nnt . 2nnt]
ry(t) = An COS + Bn Sill

and evaluate An and Bn. However, the limits on the integer n

A
have not yet been assigned. They should be 1 < n < oo,
but n would obviously have to be chosen much smaller
than the upper limit. It is hoped that for some comparatively
FREQUENCY large but finite value of n the series of the above equation
will approach the observed water surface of Fig. A-4.
Appendix A: Wave Motion as a Random Process

Figure A-4
TIME A typical wave record at sea.
0

In order to determine the coefficients An and Bn in the It is now found that

right-hand side of the above equation, it is recalled that
2 iT' 2nmt dt

i
An = - l](t) COS--
T, 2nnt . 2nmt T1 o T1
COS - - Sill - - dt = 0 for all m and n
o T1 T1 and similarly by multiplying by sin [(2nm)/T1]

i
o
2nnt
T, 2nmt
cos-- cos-- dt = 0
T1 T1
form f= n Bn = -
2 iT, 2nmt
l](t) sin-- dt
T1 o T1

form= n A-2.2.2 Sufficient and necessary conditions that the

i
function IJ(t) can be represented by this set of equations
T, . 2nnt . 2nmt d cannot be stated precisely and fully within the scope of
Sill - - Sill - - t =0 form f= n
o Tl Tl this book. (The reader is referred to many standard text-
books on Fourier analysis.) It is sufficient for the present
form= n purpose to write one of the necessary conditions, which is

To use these properties multiply both sides of the previous LT 11J(t)i dt f=

1
00

equation for IJ(t) by cos [(2nmt)/T1 ] and integrate between

0 and T1. This condition will be satisfied for wave records with T1
finite.
IT, l](t) cos--
0
2nmt
~
A IT'
dt = -2
2nmt 0
cos-- dt
~ 0
The coefficients An and Bn can be represented on two
figures such as Fig. A-3 (one for An and one for Bn). The

+ L An
n
I
T,

o
2nnt 2nmt
COS - - COS - -
T1 T1
dt
points for each coefficient will occur at multiples of kn =
(2n)/T1 where T1 is the length of the record.
In order for the Fourier coefficients An and Bn to be

+ L Bn
n
i
T1

o
2nnt 2nmt
Sill - - COS - -
T1 T1
dt
representative of a sea state for all time, the record length
T,. must be very long. The limits on n will be very large
and the practical details of such a Fourier representation
AmT1 are seen to be enormous. A comparatively short record
(n = m)
2 of the sea will yield values for An and Bn which would
 Appendix A: Wave Motion as a Random Process

only be representative of that particular record at that Hence

particular time and place. A slightly different record taken,
say, a few minutes later or a few feet away would yield A~=_!_ fT''1(t)e-inkt dt for n = 0, 1, 2 ...
completely different An and Bn. This simplified Fourier T1 o
analysis is unstable, hence a much more stable description It is seen that A~ is a complex function of frequency (nk)
of the sea state must be found. for a given record of 17(t) oflength T1 containing information
on phase and amplitude of the components making up
A-2.3 Fourier Analysis in Exponential that particular record. This operation is called a Fourier
Form transform.
The expressions for the Fourier series representation
A-2.4 Random Functions
given in the previous section will be more easily handled
in their exponential form. The transformation is as follows: There still remains the problem of deriving a more stable
description of a sea state and a method to handle longer
cos nkt = -!{einkt + e-inkt) wave records. As the record length tends to infinity, the
condition g l'7(t)l dt =f. oo will obviously be violated. The
sin nkt = ;i (einkt- e-inkt) Fourier transform giving the value of A~ will no longer be
valid. Some method has to be devised to handle a random
The previous expression for 17(t) given at the beginning process under the general concept of harmonic analysis. A
of the section may now be written: random phenomenon is one in which the fluctuations of the
quantity under observation as a function of time cannot be
A 1 00 1 00
precisely predicted. No two water level records will ever be
'1(t) = 20 + 2 ~(An - iBn)einkt + 2 ~(An + iBn)e-inkt identical. They will, however, have certain identifiable
statistical properties. In a random sea where the variety of
It follows from the definition of An and Bn that wave forms is infinite, characterization by wave form is
An= A-n since cos ( -nkt) = cos nkt contrary to the inherent feature of the process. Character-
istics that are common to all possible samples of that sea
and state are required.
Bn = -B-n since sin ( -nkt) =-sin nkt
A-2.5 Autocorrelation
The last term of the above equation can be written as Consider an observer who records a water surface
elevation at time t = t 1 at a fixed point. What can he say
about the water surface elevation at time t = t 1 + M
where M = 0.1 sec, 1.0 sec, 10 sec, or 100 sec? If the observer
Inserting this equation into the previous one gives is watching a sinusoidal wave train he can say quite a
lot about the times t 1 + l!.t but if the process is random he
L A~einkt
00

'1(t) = cannot be sure of the future. The best he can do is to give

-oo some estimate of the expected value of the sea surface
elevation. (The expected value is defined as the average
where A~ is given by
from an infinite number of observations.) The problem is to
A~ = -!{An - iBn) and A~ = !A 0 correlate the water surface elevation at time t with its
Appendix A: Wave Motion as a Random Process

value at timet + !lt. The correlation function density. The energy density spectrum really expresses
the rate of change of variance or derivative of the signal
R(r) = lim - 1 fT' IJ(t)ry(t + r) dt as a function of frequency. However, the square of the
T,-oo Tl 0 amplitude is used instead of the amplitude. The density
predicts the expected (average) value of the product of spectrum corresponding to a simple sinusoidal wave (Fig.
two values of the water surface which are separated in A-2) would appear as a spike of infinite height at the fre-
time at the same place by r. To confirm that ry(t) and quency (2n)/T since the variation of the square amplitude
ry(t + r) are taken from the same record, the expression as a function of frequency is infinity. However, the spike

s:
for R(r) is called the autocorrelation function of ry(t). has by definition a finite area:
(Occasionally the cross correlation between two different
signals may be required.) 00
S(k) dk = !A 2
Some properties ofthe autocorrelation as defined by R(r)
are developed: Such a function is referred to as a Dirac delta function. A
periodic wave of arbitrary shape which has the amplitude
1. The autocorrelation function is even, i.e., R(r) =
spectrum of Fig. A-3 will show a comb of Dirac delta
R( -r), or
functions, each one of infinite height at the frequencies
kl> k 2 , k 3 , but having finite areas equal to Ai/2, A~/2,
lim - 1 JT' ry(t)ry(t + r) dt
T,-oo T1 0 A~/2, etc.

= lim -1 fT1 ry(t)ry(t - r) dt A-2.6.2 A random function which contains no periodic

T,-oo Tl 0 components will have no infinite density spikes in the
2. The value of the autocorrelation at r = 0 is the mean frequency domain. Consider the Fourier transform of the
square water surface fluctuation since autocorrelation function

<l>(k) = -21 Joo R(r)e-ikr dr

1r -oo
3. The value of the autocorrelation as r -+oo, if the Then by definition,
observed phenomenon contains no periodic or drifting
components, is zero. The demonstration of this prop- R(r) = f_ 00

00
<l>(k)eikr dk
erty is beyond the scope of this book. It can be noted,
however, that if a process is random the correlation When r = 0,
between an observation at time t and an observation
taken at time r later would tend to become infinites-
imally small as the time r became large.
R(O) J oo
= -oo <l>(k) dk = },i~00 T
1 [T'
Jo l](t)ry(t) dt
-
= ry 2 (t)
It is seen that the total area under the curve <l>(k) is the mean
A-2.6 Autocorrelation to Energy (or
square value of ry(t). The area of any element <l>(k) dk
Variance) Density Spectrum
represents the mean square contribution of the variance
A-2.6.1 The concept of a spectrum (amplitude and phase) of ry(t) in the interval !bk. <l>(k) is the variance density
in the frequency domain can be introduced. The idea of a frequency spectrum of ry(t). It is often miscalled the energy
density spectrum is similar to the idea of a probability spectrum or the power spectrum or simply the spectrum

of the process 17(t). It can be shown that the variance
spectrum may be called the energy spectrum for deep water
linear waves. It is never a "power" spectrum as used in the
above definition, but it is always a variance spectrum for
any water surface record-linear or nonlinear.
The reason for the term energy spectrum follows from
the definition,
J~CX) <I>(k) dk = 1J 2 (t)
If IJ(t) is written as,
00
IJ(t) = L An COS (knt + ~>n)
n=l
then upon squaring and averaging the right-hand side
of the above equation
It is now recalled that the average energy per unit surface
area can be written as Eav = ipgH 2 , where His the wave
height (2An). A wave of height 2An and frequency kn in the
confused sea contributes !pgA; to the energy of that sea.
It is seen from the above equations that pg J<I>(k) dk
represents the energy in the sea (for linear waves). The
description of a sea in terms of its variance spectrum
<I>(k) dk is a purely statistical definition. If physical meaning
such as "energy spectrum" is used, it should be born
in mind that the "variance spectrum" and "energy
spectrum" are only equivalent (after a factor of pg is
inserted) for deep water small amplitude wave motions.
All the operations described above are generally per-
formed with high speed digital computers using fast fourier
transform (FFT) techniques. Detailed discussions of such
procedures are beyond the scope of this book.
A-2.6.3 Advantage can be taken of the even property
of R(r) to reduce the Fourier transform to the cosine
Appendix A: Wave Motion as a Random Process
transform. 279
<I>(k) = -1 f"'
2n _ 00
R(r) cos kr dr
and from the symmetry <I>(k) = <!>( -k) (by inspection) an
alternative definition for the spectrum is
S(k) = -1
n o
J"'
R(r) cos kr dr
If one uses the frequency spectrum in cycles/sec (Hz),
f = (k/2n), the relationship is simply:
One advantage of this definition is that the factor 1/(2n) in
the Fourier transform is not required. It is also pointed
out that in order to transform a spectrum S(k) to the
equivalent S(f) it must be recalled that this is a density
=
spectrum and the identity S(f) df S(k) dk must be
maintained so that S(k) = S(f)/2n. Some authors have
also advocated the use of" period" spectrum defined by
S(T) dT = S(f) df
where f = 1/T. This has some advantage in giving a "feel"
for the effects of wave period but in fact there is no reason
why the same "feel" cannot be developed for frequency
(for k = 2nf) once familiarity is achieved.
The spectrum and the autocorrelation function have
many desirable properties when used to describe random
processes such as sea states. They (particularly the spectrum)
have stable statistical properties in that sample records
taken from the same sea state yield closely reproducible
spectra. One further obvious usefulness is the condensation
of say a 30-min record of water surface into 50-100 points
in a frequency domain. Attempts have been made to fit
empirical curves to wave spectra.
A typical spectrum of sea-surface phenomena is shown
in Fig. A-5. This figure illustrates the complete range of the
wave spectrum from infratidal effects to capillary waves.
 Appendix A: Wave Motion as a Random Process

280
24 12 5 30 I 0.1
hr hr mrn ltl sec sec sec:

I
I ""~
!II
0
0
z., ~
cr c
Cw ltl a: ltl
-~~
g:a
ltl
""
0
ltl

""a: ""
z ""
ltl
""cr> Figure A-5
>-
C) ""
0
.....
i= C!l
::I
""C!l ..J
!II
a: -...~ ..J Energy spectrum at sea .
w """cr cr ltl
!i z
..J

z ~~ ..J
cr
z
a: ::I
a:
ltl
- ""
::I Ill
0
;;; ""3: >-
a:
w zO z
::I
0 cr 0
ltl cr
_..J
::10 a: 0 ..... z ... ..J ..J
..J
a: I ltl ::I a: II.
a:O :! ltl ::I )( iL
i
""""
:z:l-
0

""
ltl
..... ltl
"" cr
()

"""""
o:::E

10... 10-5 10- 1 10 1

A-3 Probability for Wave Motions would be (see Fig. A-6)

A-3.1 The Concepts of Probability
Distributions and Probability Density
A probability distribution is normally defined when the
proportion of values of the considered variable which are A-3.2 The Probability Density of the
less than a particular value are plotted against that value. Water Surface Ordinates
The most familiar shape for such a distribution is the typical
"S" shape, as is shown in Fig. A-6. There is no theoretical justification for the form of the
The derivative of the probability distribution is called probability density of a sea state. It is usually assumed
the probability density. The area under the probability to be a Gaussian (normal) distribution with a zero mean
density which lies between two values a, b defines the value (measured from the still water level), i.e.,
probability that the result of the event being observed lies
between a and b. The total probability of the event having
all possible outcomes must be unity.
Suppose the water surface elevation of the sea is being
considered. Then where 11 2 is the mean square surface fluctuation.
The form of this equation may be varied. If 11 is measured
in units of ('1 2 ) 1 12 ( = kq) then the above equation becomes

where p(IJ) is the probability density of the water surface

elevation '1 The probability distribution of IJ, say P(IJ),
 Appendix A: Wave Motion as a Random Process

A% 281

Figure A-6
Probability distribution offree
surface elevation.

a +
VALUE OF SURFACE ELEVATION

It is noted here that 1] 2 k;

= is the mean square surface values of kn vary from this only by a small amount, the
fluctuation and is equal to the area under the spectrum. above equation can be written:
Some authors use different definitions for the total area
under the spectrum involving factors of 2, 4, or sometimes
l](t) = L An cos (knt- kmt +en+ kmt)
8 or 16 for various reasons. Mathematically speaking, = Ac cos kmt + As sin kmt
there is a preference for retaining where

f) S(f) df = 11 2 = k; Ac
As
=
=
L An cos (knt -
- I
kmt + en)
An sin (knt - kmt + en)
and this definition is normal for spectral applications in In the equation for IJ(t) the assumption of a narrow range
most other fields. In other words, the total area under the of kn means that Ac and As vary very slowly with time.
spectrum is the variance of the process being studied. Now let R = (A? + A;) 1 i 2 . By the central limit theorem
of probability the values of Ac and As given above will be
A-3.3 Probability for Wave Heights
normally distributed so long as the sums are taken over
A one-dimensional wave train is assumed to be described a sufficiently large number of terms and
by the function
l](t) = L An cos (knt + en) A? = A; = 1] 2 = 1XlS(f) df
where the range of kn is distributed over the frequencies Therefore, the probability that Ac, As lies within the element
contained in the spectrum, the en's are arbitrary phase dAc dAs is given by the following probability density
positions and the An's are governed by the spectrum as a function
function of kn.
If the spectrum is spread over a narrow range of fre-
quencies such that the midfrequency is km and the actual
 Appendix A: Wave Motion as a Random Process

282 so long as the variables Ac and A. are statistically inde-

pendent. This follows from the fact that AcAs = 0. Now, A-3.4 Probability for Wave Periods
put
There is at present no simple development of the prob-
Ac = R cos ex A.=-Rsinr:x ability distribution of wave periods from the spectrum.
The probability of wave periods (based on zero crossings
p(A 0 A.) can be transformed by means of the identity
=
p(A 0 A.) dAc dA. p(R,r:x)R dR dr:x to
or on the crossing of 17(t) with the still water level) should be
related to the autocorrelation function in some manner
but the exact form has not been developed. In order to
1 R
p(R,r:x)R dR dr:x = - = e-
Rz;z>
dr:x dR
2n 17 z
~ introduce some estimate of the "mean period" based on
the variance density spectrum, the concept of spectrum
The variables can be separated to give moments is introduced. The moment of the spectrum of
order n is given by
p(R,r:x) dR dr:x = p(R) dR p(r:x) dr:x

This equation is in direct analogy to the moment of a

so that probability density J::' oo x"p(x) dx. The expected (average)
value for the time between successive zero crossings of a
process having the spectrum S(f) is given by

1 ~ T = ~ (Mo)l/Z = ~ ( J~ S(f) df )1/Z

p(r:x) dr:x = 2n dr:x 2 2 Mz 2 J~ PS(f)df

In fact R can be considered as the wave amplitude H/2
and ex can be considered as the wave phase position en.
The last equation indicates that the phase position of the
waves has a uniform distribution between zero and 2n.
p(R) dR can be further simplified since for a sine wave the
mean square value is equal to one-half of the square of the
amplitude.
Therefore, if H = 2R, IJz = !Rz =kHz so that
p(H) dH = 2H e- H2fH2 dH
Hz
which is the well known Rayleigh distribution.
In spite of the many apparent assumptions in the
derivation of this equation, it has been found to be an
extremely good fit for observed wave height distributions
in wind generated seas.
where Tis defined as the mean apparent wave period (the
wave period will be twice the expected time between
successive zero crossings). The proof of the above equation
can be found in more advanced textbooks and is beyond
the scope of the current chapter. In a similar manner the
expected period between successive maxima and minima
(points of zero gradient, OIJ/Ot = 0) is found to be given by
the ratio !(M 2 /M 4 ) 11 z, and this is seen to be quite different
from T/2 for a general spectrum shape S(f). The two are
identical if S(f) is considered as a delta function having
only one frequency.
The relationship of T to the autocorrelation function is
best illustrated by recalling a well-known statistical
theorem. "The moment of a probability distribution of
order n is given by the nth derivative of the characteristic
function of the probability density at the origin." The
characteristic function of a probability density is its Fourier
 Appendix A: Wave Motion as a Random Process

transform. The Fourier transform of a spectrum is the Thus,

autocorrelation function (see Section A-2.6) so that M 0 ,
M 2 , and M 4 are simply the zero, second, and fourth Su(f) df = [2nf co_sh m(d + z)J2 S (f) df
smh md ~
derivatives of the autocorrelation function at r = 0.
where Su(f) is the velocity spectrum at elevation z and
S~(f) is the surface spectrum.
A-3.5 Probability for Subsurface
Velocity and Accelerations By definition of the spectrum,

The probability distributions of subsurface velocities

and accelerations in the case of small wave motions (linear
U2 = 100
Su(f) df

assumption) are usually assumed to have a normal distri-

bution. In order to define a normal distribution it is only = Joo [2nf co_sh m(d + z)J S (f) df
2

necessary to be able to define the variance. This can be done 0 smh md ~

for velocities and accelerations in terms of the surface
It must be recalled in the transformation function that
variance spectrum and the use of hydrodynamic potential
m is a function off given by the wave equation (see Section
theory.
16-3.2)
The water surface is described by the sum of an infinite
number of sinusoids having the variance spectral distri- (2nff = mg tanh md
bution S(f). For the case of any linear random process Once u 2 is determined, the probability density distribution
it can be demonstrated that the effect of a linear operation is given by
on the process can be described in terms of the spectrum
of the process multiplied by the square of the generating 1 2
p(u) du = exp (- u ) du
transform. A sea state is described by (2nu2)lf2 uz

In a similar manner the probability density of the acceler-

ation and subsurface pressures can be derived.
The corresponding subsurface velocities and accelerations
are (see Section 16-3.4)
A-4 Discussion of Nonlinear Problems
A-4.1 Effects of Nonlinearity on
Probability Distributions
and It is well known and readily observable that high waves
tend to have shorter peaked crests and longer flatter troughs
ou(t) _ '\' 4 2f 2 A cosh m"(d + z) ( ,1 than the simple sinusoidal representation would indicate.
cos 21tjnt
)
:1
ut
- L... 1t n n . h
sm mn
d + en Since the mean water level must remain the same it would
be expected that the probability density of the surface
The square of the modulus of the operating transform for ordinates will not be the same for large sea states as for
the process u(t) is seen to be, lower ones. The positive ordinates will be larger but less
frequent whereas negative ordinates will be smaller but
[ 2nfn cos~ mn(d + z)J
2 more frequent. A skewness is introduced into the probability
smh m.d density (see Fig. A-7).
Appendix A: Wave Motion as a Random Process
NON- LINEAR (SKEWED)
The effect of skewness in water surface ordinate prob-
ability distributions is reflected to some extent in the
probability distributions of wave heights. However this
effect is not very marked and does not appear to be very
important. It must be recalled that as the crests become
more peaked the troughs become flatter. As far as wave
heights are concerned these effects are compensatory and
wave height probabilities for high seas are still very close
to the Rayleigh distribution.
The effects on wave period distributions of large seas
has not received much study. In random seas the concept
of"wave period" does not really exist. The only thing which
may be mentioned is that the times between successive
up-down crossings of the mean water level will be shorter
than the times between successive down-up crossings. The
estimates of average time between successive zero crossings
and successive maxima and minima as presented in the
section on wave period probability will not be very reliable.
A-4.2 Effects of Nonlinearities on
Spectra and Spectral Operations
Nonlinear problems can appear in wave spectrum
studies in many forms. Clearly the mechanical process
of harmonic analysis as described in Section A-2 is not
affected by nonlinear problems. It is the interpretation of
the resultant spectrum which is difficult. The high-frequency
part of the spectrum is an aggregate of the small ripples
mixed with the sea and the harmonics of some of the larger
low-frequency wave components. It would be preferable
Figure A-7
An illustration of nonlinear effects on the
probability distribution offree surface
elevation.
-- .,..,
to separate the harmonics since they travel at the phase
speed of the "fundamental" wave, whereas the shorter
waves travel much slower.
The spectrum destroys all details of phase position and
also all differences between "crests" and "troughs." The
normal wave spectrum is obviously an insufficient de-
scriptor of a high (nonlinear) sea state.
One of the major advantages of the spectral description
of a sea state became apparent in the section on subsurface
velocities and accelerations. The velocity, acceleration,
pressure, etc. spectra are easily determined from the water
surface spectrum. The transfer functions for nonlinear
sea states will not be so simple. One other form of non-
linearity arises in operations on the spectrum. For example,
the prediction of wave force spectra from sea surface
spectra when the function is of the form f = au 2 + b(oujot)
is one involving a nonlinear operation on the water surface
spectrum. These problems are not yet solved satisfactorily.
PROBLEMS
A.l Sketch the autocorrelation functions for a sine wave
and a random noise. Suggest a method for detecting a
sinusoidal component in a random process.
A.2 Sketch the variance density spectrum of a periodic wave.
Why is the description of a periodic motion in density
spectrum form not very useful?
 Appendix A: Wave Motion as a Random Process

A.3 If f(x) = x 2 and p(x) = [1/(2n) 1 i 2 ] exp -{!x 2 ), deter- Jenkins, G. M., and Watts, G. W., Spectral Analysis and Its Applica-
mine p[f(x)] and sketch. tions. Holden-Day, San Francisco, Calif., 1969.
Kinsman, B., Wind Waves. Prentice-Hall, Inc., Englewood-Cliffs,
A.4 Show that the autocorrelation function of any stationary New Jersey, 1965.
random variable is even. Lee, Y. W., Statistical Theory of Communication, Wiley, 1961.
National Academy of Sciences, Ocean wave spectra. Proceedings
A.5 In the determination of the spectrum of a random of a Conference, Basten, Maryland, May 1-4, Prentice-Hall,
process, the autocorrelation is first computed followed Englewood-Cliffs, New Jersey, 1961.
by a Fourier transformation in the form of a cosine trans-
formation. What would you expect to find if a sine trans-
form on the autocorrelation was performed? Why?
A.6 A random variable x has the exponential probability
density
p(x) =a exp ( -blxl)
where a and b are constants. Determine the relationship
between a and b and the probability distribution function
P(x). Sketch p(x) and P(x).

A.7 On the assumption of linear wave motion, derive the

function for the variance of the subsurface bottom pressure
in water depth d for a sea state given by
l](t) = L An cos (nfnt + an)
A.S The probability density distribution of wave heights
in a confused sea is given by

p(H) dH = H exp (
2H 2 H2
H2) dH
Establish the relationships between the most probable
wave height, the average wave height, the significant
wave height (the mean of the highest one-third of the wave),
and the highest wave (probability 0.01).

REFERENCES FOR APPENDIX A

Blackman, R. B., and Tukey, J. W., The Measurement of Power
Spectra. Dover Publications, New York, 1958.
Cosley, J. W., and Tukey, J. W., An algorithm for the machine
calculation of complex Fourier series. Math. Computation,
Vol. 19: 90. 297, 1965.
 Appendix B
286
B-1 Introduction to Scale Model
Technology

B-1.1 Definition and Advantages of

Similitude and Scale Models
Scale model technology is based on similitude or similar-
Scale Model Technology ity between two phenomena at different scales. Similitude
is a science which is aimed at establishing general functional
relationships between various parameters based on di-
mensional analysis. Scale model technology is an engineer-
ing art in which a compromise is made with the law of
similitude, to build a practical tool with which engineering
problems can be solved.
If problems are too complicated to be analyzed by
application of theoretical methods, the answer is scale
model technology. The advantage of scale model vs theory
is threefold:

1. Complex boundary conditions cannot be analyzed

by analytical means. Numerical solutions and high-
speed computers extend the power of analytical
methods used. But scale models are the best analog
computer.
2. Nonlinear effects are still the greatest source of
mathematical difficulty. It is true that many problems
can often be linearized, but the great advantage of
scale model study is the reproduction of not only the
linear forces in similitude, but also the nonlinear
convective inertial force pV(V 2 /2). The convective
rotational force pV x curl V may not always be
reproduced in similitude, as the distribution of
rotationality curl V is often related to the distribution
of friction forces, which cannot be reproduced in
similitude exactly. This scale effect is often negligible,
as will be seen in the following.
3. Short fully turbulent ftow are statistically in similitude.

In general, a scale model does not reproduce all the

aspects of the phenomenon under investigation, but a few

effects which are of interest in the research. A knowledge
and understanding of the physical laws which govern the
effects studied permits us to determine the relative influence
of" scale effects."
B-1.2 Economic Considerations
The scale is chosen as a compromise between economics
on the one hand and the technical requirements for simili-
tude on the other hand. From the economical viewpoint,
the cost of scale model experiments increases approxi-
mately as the cube of the scale A3 , since the area of the model
increases as A2 , and the theoretical duration of the test
increases as A112 , but the actual duration increases faster.
Of course, this rule of thumb has many exceptions and is
not valid if the scale of the model is too small. (In the case
of a very small model, the cost of the operation actually
increases as the difficulty of carrying out reliable measure-
ments increases.) The "best" scale is not the largest possible
one. As a matter of fact, the scales of models must primarily
be dictated by economical considerations. Actually, the
decisive elements rarely present themselves in a precise
manner. It usually is almost impossible to evaluate, even
roughly, the "economic expectation" of a model and
particularly the variation in this economy with respect to
the accuracy of the tests. Hence, the determination of the
scales of optimum reduction will necessarily be an art
rather than a science, but the value of the results obtained
will nevertheless primarily depend upon the overall know-
ledge available to the designer. Among other things, it is
imperative that he be acquainted with the laws of similitude
because not only will they enable him to estimate the degree
of precision that he may expect from the model, but, above
all, they will render it possible to have a sound concept of
the latter and to achieve the optimum precision compatible
with a given sum available for research.
The first stage of the study of similitude is not concerned
with a detailed discussion of the equations, but is essentially
an examination ofthe problem as a whole and of its physical
character. For example, it is impossible to form a sound
Appendix B: Similitude and Scale Model Technology
concept of a model without a priori having an idea of the 287
causes and origin of the phenomenon studied. The overall
examination of the phenomena involved enables us, a
priori, to eliminate basically erroneous methods of ap-
proaching the problem. It renders it possible to submit a
broad outline of the boundaries of the rational conditions.
In many cases, it also will enable us to determine the princi-
pal stages of the study and occasionally, the number and
type of the models required. The detailed study of the
equations ofthe phenomena will then define the approxima-
tions, scales, distortions, etc., and eventually, the number of
models required, as well as the rules of similitude to be
applied to each of them. For example, in the case of a
conventional study of the protection of a port against
wave action, it may happen that the installations as a
whole are studied on a scale of 1/150, some particularly
important structures (such as those at the entrance) on a
scale of 1/75, and the stability of the individual structures
on a scale of 1/40.
In summary, the problem of similitude is far from simple,
and there are no standard solutions. The engineer has at his
disposal a number of complex means which he must be
able to handle with skill. The better he is acquainted with
the tools, the more effective and economical the solution
will be. From the technical viewpoint, the problem consists
of examining when scale effects become important. This
leads us to consider the forces which always need to be in
similitude. Each separate case needs to be examined in
detail. However, some general rules can be presented.
For all practical purposes, a scale model must fulfill the
following conditions:
1. It must be exact; i.e., it must reproduce with exactness
the natural phenomenon under study.
2. It must be consistent; i.e., it must always give the same
results under the same conditions.
3. It must be sensitive, or more exactly, its sensitivity has
to be imposed by the needs of the reproduction of the
phenomenon under investigation.
4. It must be economical, of reasonable size, and
completed within a reasonable time interval.
 Appendix B: Similitude and Scale Model Technology
288
B-1.3 General Considerations on the
Concept of Similitude and Scale Model
Technology
B-1.3.1 The rules of similitude can be obtained by three
different approaches: dimensional analysis, inspectional
analysis, or the most general method, which consists of
deducing the conditions of similitude from the Navier-
Stokes equation and the equations of elasticity. In text-
books on hydraulics, similitude is usually presented as a
natural consequence of dimensional analysis. Then the
similitudes of Froude, Reynolds, Mach, Weber, Cauchy,
etc. are presented. But no scale model has ever been built
according to an equality of Weber number, or even of
Reynolds number. Practically, a Reynolds similitude does
not exist in scale model technology; but the similitude of
head loss, a function of the Reynolds number, is sometimes
adjusted.
Because of the inherent inadequacies of dimensional
analysis, the engineer will often rather deduce the laws of
similitude by "inspectional analysis." Knowledge and
understanding of the phenomena under study is necessary
for deducing the rules of similitude, for neglecting phenom-
ena of secondary importance, and for deciding the relative
importance of scale effects. Inspectional analysis is also
required for the interpretation of scale model results. The
problem may need to be mathematically formulated in a
differential form, but does not need to be integrated. The
law of similitude can be deduced from the law of motion
under a differential form.
B-1.3.2 The dynamics of a system obeys the general law
I+G+P+F+E+C=O
where I is the inertial force, G the gravity force, P the
pressure force, F the friction force, E the elastic force, and
C the capillary force. These forces could be expressed
vectorially, or on projection axis, in differential form, or in
an integrated form. On the scale model, one has:
I' + G' + P' + F' + E' + C' = 0
and for similitude
I/1' = G/G' = = C/C'
Once all these factors are mathematically expressed, the
conditions of similitude are obtained. It is then realized that
complete similitude is impossible. Hence, it is important to
know what can be neglected and what are the scale effects.
This is the inspectional analysis method.
B-1.3.3 It is then important to distinguish between
conditions of similitude and criteria of similitude. The
conditions of similitude are an ensemble of formulas
deduced from the physical laws governing the phenomena
under investigation; e.g., the similitude condition governing
sediment transport is obtained by an analysis of the
mechanics of sediment transport. They have an absolute
definition which cannot be changed unless an improvement
in the knowledge of the physical law is obtained. They are
not chosen by the experimenter, but are imposed on him.
Unfortunately, it is known that in the field of sediment
transport, many phenomena still remain to be analyzed or
clarified. Thus, the conditions of similitude are not as well
defined as they should be. A choice of what is important
will have to be based again on the knowledge of these laws
obtained by "inspectional analysis."
In performing a model study, an experimenter must
specify certain criteria, such as model wave conditions and
fineness of model bottom features. The criteria of similitude
is a free choice of the experimenter to a very large extent.
For example, sea states vary from day to day and from hour
to hour. The experimenter will choose, for the sake of
simplicity as well as practical necessity, a characteristic
wave condition and will only be able to reproduce simplified
storm and swell conditions on the scale model. He will
choose the wave direction and the wave amplitude and the
duration guided by his knowledge of natural conditions.
In particular, the wave generator will generate waves at an
angle which corresponds to the dominant direction of
storm wave energy. Even though he is guided by his knowl-
edge of wave statistics, his final choice will be determined
by a trial and error method which permits him to reproduce,

for example, a wave agitation pattern or the same bottom
evolution as observed in the prototype. This faithful
reproduction of bottom evolution determines the choice of
the wave characteristics, rather than the strict conditions
of similitude of wave motion.
Other criteria of similitude will be to what extent he
wants to reproduce the fineness of the bottom topography, a
typical tide cycle, a typical flood wave, the currents and
their variations with time, and so on. In summary, the
criteria of similitude are specified by the experimenter as
reasonable approximations for simplification of model
operation.
B-2 Short Model vs Long Model
B-2.1 Froude Similitude: Nondissipative
Short Models
B-2.1.1 There are two kinds of scale models for which the
law of similitude are deduced: short models and long
models. The short models are considered either as non-
dissipative or as fully turbulent. The nondissipative short
models are the ones where the flow pattern is essentially
governed by inertia force and gravity only in the case of free
surface flow, or inertia forces and pressure gradient only
in the case of flow under pressure.
The ratio of inertial force (dimensionally equal to p V 2 ),
to gravity or pressure force (dimensionally equal to pgL) is
a Froude number V 2 j(gL), which is to be the same on a
scale model as on a prototype. The Froude similitude stems
from the fact that the gravity acceleration is the same on
the model as in the prototype. An example of such a scale
model is the flow over a spillway, when boundary layer
effects are negligible.
The inertial force and the gravity and pressure forces are
always present in water wave problems, and the Froude
similitude applies when viscous effects are negligible.
B-2.1.2 From an equality of Froude numbers,
vz'
2
gL
y 'P = gL m which the air bubble penetrates in a breaker will be relatively
Appendix B: Similitude and Scale Model Technology
(p refers to prototype, m to model) one deduces
( Vm)2
vp
= Lm =A
Lp
where A is the geometric scale. Thus the ratio of scale veloc-
ities is A112 . The ratio for time scale is also A112 , since
Tm
Tp
The ratio of discharges is like the ratio of areas A2 times
ratio of velocities A112 ; i.e., A5 12 . The ratio of powers is
discharge times height, i.e., A7/2. The ratio of angular
velocities or frequencies 1/T is A- 112 , etc. Since the ratio of
pg is unity, the ratio of pressure is A, and the ratio or scale
of forces is (area x pressure) A3 . The ratio of powers
defined by force times velocity is A712 . The ratio of shearing
stress r is also A like the pressure.
B-2 .2 Generalized Froude Similitude:
Fully Turbulent Short Models
B-2.2.1 Let us now consider the viscous forces. It is
known that the ratio of inertial force to viscous force is
dimensionally represented by a Reynolds number, VL/v,
where vis the kinematic viscosity. An equality of Reynolds
number and Froude number is possible only at scale unity.
Hence, Froude similitude is possible in two cases only:
(1) where viscous forces are negligible, as in a gravity
wave prior to breaking or in a flow over a weir; or (2) where
the flow is very turbulent and the flow pattern to be repro-
duced on scale model is short, as in a hydraulic jump or a
breaking wave. Indeed, in the latter case, the dissipation of
energy is mostly due to turbulent fluctuations and is not
due to laminar viscous effect. While these viscous effects are
linearly related to the velocity, the turbulent fluctuations
are quadratic, i.e., proportional to the square of the average
velocity V 2 , as are the inertial forces. Thus, the ratio of dissi-
pative forces to gravity force in a very turbulent flow is also
a Froude number. This situation allows us to use the
"generalized Froude similitude." Of course, the depth at
Appendix B: Similitude and Scale Model Technology

larger in the field than in the model, as the size of the bubble
(determined by capillary effects) is approximately the
same in any air entrainment phenomenon. But the dissipa-
tive forces are in similitude and proportional to V 2 , and the
total amount of energy dissipated is in similitude (A 7 12 ),
even though the fine structure of the flow could be different.
B-2.2.2 This is evidenced when one considers, for
example, the shearing stress (see Section 8-3.3):
- au (au) 2 au
r = pu'v' + 11 ay = p/2 ay + 11 ay
where uis an average velocity, 11 the viscosity, p the density,
l the Prandtl mixing length, and u' and v' the turbulent
velocity components. In the case where rm = A:r P and
which is in accordance with the Froude similitude, since
the rate of energy dissipation is a power dimension.
So a small hydraulic jump, defined by h 1 , h2 , and Q, is in
similitude with a large jump defined by h 1 = Ah'1 , h2 =
Ah~, and Q = A512 Q' at a scale A. Indeed, the phenomenon is
very turbulent. The average flow pattern is the same. The
turbulent fluctuation may statistically be different, but the
gross effect remains the same. It is the reason why stilling
basins, for example, could be investigated by scale models.
In the case of breaking waves, it is sufficient to remember
that scale effects will be negligible for the phenomena under
consideration, provided the height of breaker is larger than
say, 5 em. This statement is based more on experience than
on theory.

a- ~ p [2
11 __!!_
ay
(a-)__!!_
ay
2
B-2.2.4 From the most general viewpoint, it is known
that the average fluid flow obeys the Reynolds equation
au; -au; a(p + pgz) a 2 U; a ( -,-,)
it can easily be seen that the shearing forces r will be in p-a
t
+ puj-a
xi
= a
X;
+ 11 aX;- a
xi
--a pu;uj
xi
similitude provided the mixing lengths are such that l' = AI
in accordance with the similitude of Froude. This condition where (U;) is the average velocity and u; is the turbulent
is fulfilled in the case of fully turbulent motion occurring fluctuation velocity. It is known that u; = u; and u; = 0 in
over a short distance and presenting a large velocity the case of a perfect fluid (11 = 0) and in the case of a laminar
gradient. flow. If this equation applies for the prototype, a similar
equation should also apply for the scale model, and for
similitude one should have
B-2.2.3 An example will illustrate these considerations.
The rate of energy dissipated by a hydraulic jump or a _au; I
pui- -I
a(pgz)
-
moving bore, dE/dt, depends only upon the discharge Q axi m ax; m pu;u}lm
and the depth of water before and after the jump (h 1 and h2 ). -au; I = ... = a(pgz)l - pu;ujlp
One has seen (Section 14-5.1.2) that the rate of energy dissi- pui- --
axi p ax; p
pation by an hydraulic jump is:
dE _ Q (h 1 - h2 ) 3
=
Since Pm = pP, g g and zm/zP = A, one may deduce that
the scale for pressure p is also A. The scale for acceleration
dt - pg 4h1h2
au;/at is unity, as is g. Consequently, the scale for time is
independently of the scale. Therefore, the ratio A112 , and the scale for velocity is

dEl ~~m = Al/2

dt m = AS/2 A3 = A?/2
u;!P
dEl A2 It is seen also that a similitude of viscous force is impossible
dt p unless A = unity, and that a similitude of Reynolds stresses
 Appendix B: Similitude and Scale Model Technology

is possible only if u;uj :::::; (~) 2 which is the case of very
turbulent flow on short structures. It is also seen that
similitude is possible, as previously stated, in the case of a
perfect fluid (f.l = 0, il; = u;), where the only forces are
inertia (linear or nonlinear) and gravity-pressure. Con-
sequently, a first scale effect is due mostly to the fact that the
viscous forces are not in similitude. The value of scale model
experiments depends upon their relative importance, i.e.,
the thickness of the boundary layer and its effect on related
phenomena such as separation and wakes. Fortunately,
many fluid flows may be successfully considered as that
of a perfect fluid (f.l = 0) or as very turbulent. The Froude
similitude or generalized Froude similitude is then the rule
for all short flow patterns.
B-2.2.5 In the case of flow under pressure, which could
be subjected to cavitation the similitude parameter
11pj(p V 2 ), may intervene requiring to reduce the atmospheric
pressure Pa to a scale model partial vacuum.
B-2.3 Similitude of Head Loss; Long
Models
the inertial forces. This is the generalized Froude similitude 291
seen in the previous section.
On the other hand, in a long model, friction has a
definite influence on the flow pattern; therefore, in addition,
a similitude of head loss is required. This head loss is a
function of the Reynolds number, but is not determined by
the so-called Reynolds similitude requiring an equality of
Reynolds numbers. Therefore, similitude for long models
requires, in addition to the Froude similitude, another
condition which makes the similitude sometimes impossible
as it is shown in the following section.
B-2.3.2 Indeed, the head loss /1H in a pipeline is
11H _ (UD) U 2
l - f v 2gD
where lis the length of the pipe, Dis its diameter, and U is
the average velocity. For similitude of inertia and gravity,
one must have
uzl uzl
gD m = gD P

B-2.3.1 It has been seen in elementary hydraulics that

and for geometric similarity:
one has Froude similitude, defined by an equality of
Froude numbers, and the Reynolds similitude, defined by
an equality of Reynolds numbers. Actually, very few scale
;).HI
l
=;).HI
ml p
models are ever built according to an equality of Reynolds
number, as the fluid velocity will have to be larger on the Therefore,
scale model as it is on the prototype (unless one uses dif-
ferent fluids).
For practical purposes, to the old teaching tradition
which consists of presenting a parallel between Froude and This cannot be achieved in the case of a smooth prototype
Reynolds similitude, one will make a parallel between what pipeline. Indeed, the similitude for Reynolds numbers is
we can call "short model" and "long model." In a short-
scale model, viscous friction is unimportant as compared UDI = UDI X A_3/2
V m V P
to gravity and inertia; therefore, it is governed by Froude
similitude, as previously seen. Also, energy dissipation may i.e., (UDjv)lm < (UD/v)ip, and fm > fp The friction co-
result from a fully turbulent condition, as in the case of a efficient fm can only be larger than JP, as shown in Fig.
hydraulic jump or a wave breaking on a beach. Boundary 13-16 or in Section 13-4.3.2. In the case of a rocky rough
layer effects in both cases are unimportant. The dissipative gallery (not lined by concrete walls), the relative roughness
forces are also proportional to the square of velocity like is large and the boundary layer is turbulent. Then, it is
Appendix B: Similitude and Scale Model Technology
possible to adjust by trial and error the scale model rough-
ness in order to obtain the same friction coefficient. A
"Froudian" discharge is then obtained, i.e., Qm = QPA 5 12 ,
when the relative roughness
are theoretically identical, and the Reynolds number
(U Djv) lm exceeds a given value, in such a way that the
boundary layer is "completely rough," or again k 5 U*jv
> 70 (see Fig. 13-15). But, in any case, the velocity distri-
bution which is also a function of the Reynolds number
(see Fig. 13-13) could never be in similitude exactly. It is a
scale effect, which is most often acceptable. The same
considerations apply to free surface flows.
B-2.3.3 Water wave scale models could, in general, be
considered as short models, whether in Froude similitude
(wave propagation) or in the generalized Froude similitude
(wave breaking). The magnitude of long shore currents and
location of rip currents may, to some extent, depend upon
friction characteristics of the beaches, in which case the
study of this phenomena would have to be considered as
belonging to the categories of long models and, therefore,
may not be studied on scale models. Nevertheless, most
scale model studies have to deal with short coastal structures
(like entrance of harbor); thus, the water motion is not too
dependent upon the friction coefficients. The main dissipa-
tive mechanism is due to wave breaking. If viscous damping
is too significant, as is the case when the wave has to travel
a long distance in very shallow water, a correction co-
efficient, determined theoretically, can be applied. Very
rarely do we have to be concerned with adjustments of
roughness for similitude of energy dissipation, and coastal
models can be considered as short models. On the other
hand, as has been pointed out previously, a similitude of
head loss is imperative for models of rivers and estuaries-
these are long models. Despite this adjustment, the vertical
velocity distribution which is a function of the Reynolds
number can never be in similitude.
B-2.3.4 Ship resistance is due to the waves created by
the hull displacement, and to the boundary layer drag and
flow separation. The first phenomenon-the ship wave or
Kelvin wave-would exist even in a perfect fluid, and it is a
function of a Froude number only. From this standpoint,
the ship is a short model and this effect could be investigated
in similitude. But the boundary layer development along
the hull is a long model, and since the hull is smooth, this
force is not reproduced in similitude. One can only obtain
an approximate effect, introducing a thin wire at the bow
of the scale model, in order to cause some turbulence in the
scale model boundary layer. Similarly, waves breaking on
the large armor blocks covering a breakwater core create
very violent and turbulent flow motion around these
blocks. Therefore, it is to be expected that the corresponding
pressure forces will be in similitude, and studies on stability
of breakwater are done according to the law of the general-
ized Froude similitude (a short model). The study of the
permeability of breakwater requires the reproduction of
similitude of permeability and energy loss by distorting
the size of the material of the core of the breakwater. From
this standpoint, these are long models, since the stream
tubes within the porous core are long.
B-2.4 On Distortion
B-2.4.1 A model is distorted when one of its scales (say
the vertical scale Jl) is different from the horizontal scale A.
D = 11/A is the rate of distortion. A long model can also
have two vertical scales, one for cross sections and another
one for the bottom or energy slope, For example, the slope
of a scale model river could be adjusted to satisfy a similitude
condition for head losses.
B-2.4.2 A short model cannot be distorted as the flow
pattern on the scale model would be completely different
from what it is on the prototype. A long model could be
distorted to some extent, without too much scale effect. It is
actually a method frequently used to overcome the con-
straint imposed by the required equality of friction co-
efficient between model and prototype. Also, it permits
 Appendix B: Similitude and Scale Model Technology

reduction of the horizontal scale of the model (i.e., the cost) B-2.4.5 In the case of a movable bed scale model which is 293
without sacrificing too much accuracy. used to study sediment transport and bottom evolution of
rivers, estuaries, and beaches, distortion is not an engineer-
B-2.4.3 In gen~ral, scale models used in the study of ing trick for reducing the size of the model and the bottom
water waves cannot be distorted. The wave velocity is friction, but is the extrapolation of a natural observed
phenomenon. The method to obtain a satisfactory scale
gT 2nd model is first to obey the law of nature, even though this law
C =-tanh~
2n L may not be fully understood. For example, a small river
The tanh lm must be the same as the tanh IP and this is only flowing in its own alluvium can be considered a distorted
possible if model of a large river. This means that the ratio depth to
width of the small river is comparatively greater than the
relative depth of the larger one. The ratios of depths J1 and
widths A are approximately related by the law of Lacey
(A 2 = J1 3 ), in accordance with the" regime theory." Similarly,
Then both time scale and wave velocity scale are like A112 a beach in a protected area has a relatively steep slope,
Since the ratio of wavelengths L,JLP is given by the hori- whereas a beach in an exposed area tends to have a more
zontal scale A, it is deduced that the ratio of water depth gentle slope. The vertical scale being defined by the ratio of
dm/dP should also be A. Then the wave pattern is in similitude. incident wave heights, a protected beach can be considered
As such, water wave models could be considered as short as a distorted scale model of an exposed beach. In both the
models despite some viscous damping. There are some case of the river and the beach, the choice of distortion
exceptions in the case of long waves and in the case of becomes a stringent condition to be respected quantitatively.
movable bed scale models. However, the natural law determined by statistical observa-
tion of natural phenomena needs to be modified in the case
B-2.4.4 In the case of long waves of a river model because scale models generally do not use
2nd the same material as the prototypes in order to satisfy
h 2nd
tan~~~ other conditions of similitude. In the case of beaches, a
L L
natural distortion based on the use of sand both in the
and scale model and in the prototype is unavoidable. However,
c= (gd)1f2
the use of sand is not always recommended because it leads
L = T(gd)1f2
to too large a distortion and, subsequently, to large scale
The model could then be distorted. The velocity scale is effects. Lighter materials (pumice, coals, plastic) are
preferred.

B-2.4.6 Similitude of wave refraction is also obtained

when dm/dp = J1 and L,JLP = Jl, which is obtained with
and the time scale is
Tm/TP = J1 112 However, the model could still be distorted.
Such similitude is used in movable bed scale model of
beaches.
It is pertinent to point out that since the only require-
This time scale also applies to a distorted tidal model and a ment of a movable bed model is a reproduction of bottom
distorted model of a river. evolution, it is not necessary that this be achieved through
 Appendix B: Similitude and Scale Model Technology

294 exact similitude of water motion. Since the model is After some straightforward calculation, one obtains the
distorted, similitudes of wave refraction and wave breaking scale effect defined by 11L/L from the following equations:
are only aimed at. These conditions are the most susceptible In the case where d/L is small,
to producing satisfactory long shore current and sediment
!1L 1.55 x w- 3
transport distribution. This is achieved by keeping the
L~ dT 2
ratio of wave lengths and wave heights like the vertical scale
Jl. Based upon this condition, the following wave character- where dis in centimeters and T is in seconds. If (2nd)/L is
istics are preserved in the model: (1) wave steepness; large,
(2) refraction pattern and angle of refraction with bottom
!1L ys
contours; (3) breaking angle of wave crests with shorelines, if
the distortion is not too large; and (4) breaking depth. T ~ (0.105)4
Also, the scale for long shore current and mass transport Figure B-1 illustrates these results, where 11L/L is given as
velocities is approximately J1 1 / 2 . Therefore, the ratio of a function of T and d. It is seen that 11L/L is smaller than
scales of wave particle velocity to current velocity is 1 %when T > 0.35 sec and d > 2 em. In practice, it can be
approximately unity. The study of movable scale model stated that the lower limit of limiting scale effects is
technology is beyond the scope of this book. T > 0.35 sec.

B-3 Scale Effects Figure B-1 Capillary effects.

B-3.1 Capillary Effects

The influence of capillary effects has to be minimized on
scale models of water waves for two reasons:
1. The wave celerity is a function of surface tension.
2. Surface tension introduces a wave damping effect.
Let us consider the first problem and try to determine
the suitable minimum wave period for scale experiments. It e
is known that the wave velocity (first order of approxima-
tion) is given by the following formula (Problem 16-17):
...X"- l~--~r4--+-~------~----------------~
1-
Q.

cz = (gL' + ~ 2n) tanh 2nd

11.1
0

2n p L' L'
where L' is the wave length (C = L'T), T the wave period,
g the gravity acceleration, d the water depth, p the density,
and A the surface tension (74 dynjcm for air-water interface
at 20C). If L is the wavelength when A is assumed to be
equal to 0, then

PERIOD T, sec
 Appendix B: Similitude and Scale Model Technology

weight W of the rocks or armor units covering the break- 295

water is proportional to
B-3.2 Wave Damping
Let us now consider the wave damping phenomenon by
viscous effects. The wave damping phenomenon is strongly
affected by scale; viscous damping is mostly due to bound-
W ~
[(Pb cos IJ. -
PbH3
;:n
Ps) A - Pb sin IJ. J
ary dissipation. The wave height H is a function of the hori- where a is the slope of the breakwater near the free surface
zontal distance x as and tan A ~ 1. Pb is the density of the armor, and Ps the
density of sea water; ). is a friction angle for the material.
This formula can be approximated as

"' ~ (
where L\ is the damping coefficient.
In a rectangular tank of width b and depth d gp)~'
Pb;. Ps K!i cot IJ.
L\ = L\b + L\s
L\b is due to the solid walls and is as presently used by engineers. Ka is a damage coefficient
~ 3 for rocks. Therefore, for similitude:
L\b =2m (~) 112 mb + sinh 2md
b 2k 2md + sinh 2md
where m is the wavenumber and equals 2n/L, k is the
frequency and equals (2n)/T, and vis the kinematic viscosity.
L\ 5 is due to a contaminated film on the free surface which,
at the maximum, can be where p is the fresh water density. It is found that

L\
s
= 2m (~) 112 mb cosh 2 md
b 2k 2md + sinh 2md
From these formulas, it can be verified that the viscous
damping is not negligible in many wave tanks where the
water depth is smaller than 2 em. This water depth should
be considered as a reasonable lower limit for scale model
studies.
B-3.3 Density Effects
One important scale effect is due to the fact that in the
wave tank one usually uses fresh water instead of sea water.
The slight difference in density, which is approximately 3 %,
changes the wave forces accordingly. Let us consider, for
example, the case of a rockfill breakwater. The minimum
(p,,:~ p')' ~ (p'~~ p)'
If Pbm = Pbp it is seen that the error on W due to the use of
fresh water is about 10%.
B-3.4 Wave Forces on Structures
The important case of a cylindrical structure is now
analyzed. One must distinguish between large piles and
small piles, characterized by the value of the ratio of the
diameter D to the wavelength L: D/L. In the first case
(small D/L), the wave force is given by the Morison formula
as a function of the horizontal component of velocity u
of the incident wave as (see Section 13-5.2.8):
dF = (PcvD u; + pCM n~ 2 ~~) dz
Appendix B: Similitude and Scale Model Technology
where ev is a drag coefficient and eM is the virtual mass
coefficient. ev and eM are supposed to be constant param-
eters which are functions of the Reynolds number, (uD)jv,
roughness of the pile, etc. Actually, ev and eM are time
dependent functions (but these functions are still not well
defined), and the wave velocity field u is not free, but
modified by the presence of the pile. The coefficients ev
and eM are obtained experimentally. The coefficients eM
and ev must be the same on the model and on the proto-
type. For a perfect fluid, eM = 2 and ev = 0. In practice, a
wake develops on the lee side and on the front side of the
pile alternately, hence eM # 2 and ev > 0.
The wake effect does not change the value of eM very
much, but it strongly affects the drag coefficient ev. The
dependence of ev upon the Reynolds number makes simili-
tude possible only if (uD)/v > 2 x 10 5 in the case of
uniform flow, but actually u is time dependent and varies
from positive value to negative value alternately. The effect
of roughness remains small, except that it may change the
inception of flow separation and hence influence the wake.
From this standpoint, the development of the boundary
layer around the pile is to be considered as a long model
(even though the body itself is short). Therefore, wave
forces on small piles cannot be studied accurately on scale
models unless the ratio of inertial force to drag force is
large. This ratio is represented by the Iversen modulus
aul D
ot max
U 2 1max
which is actually an inverse Froude number where the
gravity acceleration g is replaced by the particle accelera-
tion oujot.
In the case of very large piles (large D/L) the problem is
one of wave diffraction. The drag force is small, and in some
cases (like large circular or ellipsoid cylinders), the problem
can be linearized and treated completely analytically with
good accuracy. The solution is given by a potential function.
It is evident that such wave motions around very large piles
(D/L > 0.5) are very well reproduced on scale models even
at a small scale. Moreover, complex boundary conditions
due to complex structural forms and nonlinear effects,
particularly important in the case where the wave breaks
on the structure, are also in similitude, whereas they cannot
be analyzed by theory.
All these considerations being borne in mind, it is pos-
sible to establish the critical value of H/D for which inertial
forces
(F M)max = peM nD 2
-
4
J~ (au)
;;- dz
-d ut max
equal the drag force
(F v)max = pevD J~ U~ax dz
-d
as a function of the wave height H and the relative depth
D/L. u, at a first order of approximation, is equal to:
_ k H cosh m(d + z) (k _ )
u - 2 . h d cos t mx
sm m
The calculations have been performed for two values of the
wave height, namely H ~ 0 and H = Hb, such that Hb/Lb =
0.14 tanh (2nd/Lb) (limit wave steepness).
For the sake of simplicity, this work has been carried
out by assuming ev = 1 and eM= 2 as the most realistic
average values. The results of these calculations are
presented in Fig. B-2. On the high side of these two curves,
the drag forces dominate and there is no way of obtaining a
satisfying scale model investigation except at scale unity.
On the low side, the inertial force dominates in such a way
that similitude is more valid.
From these considerations, it is seen that in the case of
small D/L, the larger the scale, the better. However, large
values of D/L can be studied satisfactorily even at a small
scale.
This nomograph can also be used as a guide for forms of
body other than circular pile. For example, if D is con-
sidered as the beam of a ship, it is seen that the study of ship
behavior in water waves can be done in similitude.
 Appendix B: Similitude and Scale Model Technology

The Cauchy similitude is obtained either from the 297

Boussinesq equation for elastic material or from the
equation of the phenomena to be studied, through in-
spectional analysis. In general, it is sufficient that the Poisson
ratio be the same on the model as in the prototype, and the
modulus of elasticity Em be such that Em = EPA..

B-3.5.2 Similarly, in studies of stratified fluid (thermal

outfall), the prevailing conditions of similitude are imposed
by an equality of Richardson number and densimetric
Froude number.
The Richardson number is dimensionally the dynamic
ratio of the buoyancy to inertia force gradient and is a
direct measure of the stability of a density stratified flow.
VIRTUAL MASS FORCES
c
.....
PREDOMINATE g dp
:r (IN SIMILITUDE l
p dz
R; ~ (~:)'
pis the density, z a vertical coordinate, and u the horizontal
velocity. The densimetric Froude number is
0.1 L---'--'---'--.L...I....I-I-'------'---'----'---'---'-.J....J....L...I
0.02 0.1
d/L DEPTH /WAVE LENGTH uz
Fr=~
Figure B-2 Forces on cylinders.
-gd
p
Another significant criterion used in engineering practice where dis a vertical distance (depth) and 11p the variation
is the "excursion ratio"-the ratio of the horizontal wave of density about ambient. Then the ratio of Richardson
particle amplitude to pile diameter, which for the sake of numbers
similitude has to remain smaller than unity.

B-3.5 Special Effects

B-3.5.1 The Cauchy similitude, or similitude of elasticity,
could be made compatible with the Froude similitude.
This similitude is required to study ship mooring, for
example, since the elastic forces due to mooring lines have
to be in similitude with other forces. (Also, the solid friction
of the ship against the fenders has to be made compatible.
This is done by making the fenders very slippery, smooth,
and oily.)
Thus, preserving the densimetric Froude number also
preserves the Richardson number.
One also finds a short model (near the outfall) which
cannot be distorted and a long model for the study of the
far field which could be distorted. Then the similitude of
surface heat loss imposes some scale effects, the treatment
of which is judged beyond the scope of the present book.
Appendix B: Similitude and Scale Model Technology
B-3.6 Other Scale Effects
Let us just mention:
1. The solid friction forces are generally relatively larger
on scale model than on prototype (i.e., a concrete
cap on rockfill breakwater or a ship against a fender).
2. The structural strength of material is generally not
reproduced in similitude. For example, if the armor
units covering breakwaters were in complete struc-
tural similitude, one should be able to crush them in
the hand. Scale models of Dolosses or Tetrapods,
scaled at 1/50, falling from a height of, say, 1 m, do
not break, whereas their prototype counterparts
falling from 50 m certainly do.
3. Air compression effects, such as due to the slamming
of a wave against a vertical wall or buoyant tanks
subjected to pressure variations cannot be investi-
gated in similitude, unless the scale model is operated
in a partial vacuum.
4. Air entrainments such as found in hydraulic jumps or
in a breaking wave do not present the same con-
centration distribution at a small scale as in the
prototype.
5. The wave diffraction pattern on a distorted model is
not in similitude. Similitude of turbulent fluctuations
and sediment transport by suspension does not exist.
Only a similar bottom evolution could be attempted
after a lengthy calibration of the scale model based on
prototype observations.
6. Shock waves due to underwater explosions are not in
similitude since the compressibility of water is not
scalable. The same reasoning applies to slamming or
impact phenomena. The free surface is "harder" on
scale models.
7. The wave reflection coefficient of a smooth wall is
smaller on a scale model than on the prototype. The
opposite tends to apply in the case of wave absorbers
made of ripraps (large rocks). This scale effect is
corrected by adding a wire mesh on the scale model
wave absorber.
B-4 Typical Scales Used in
Engineering Practice
1. Breakwater stabilityjrockfill cofferdam: 1/30 to 1/50.
2. Wind-wave penetration in harbor: 1/100 to 1/150.
(The water depth cannot be smaller than 2 em and
the wave period has to be larger than 0.5 sec)
3. Spillways, bottom outlet, water power structures:
typically, 1/50 to 1/100.
4. River, estuary-distorted: typically, 1/100 vertical;
1/800 horizontal.
5. Beaches, shoreline processes-distorted: typically,
1/100 vertical; 1/300 horizontal.
6. Ship dynamics problems: typically, 1/100.
PROBLEM
B.1 One wants to reproduce a permeable rockfill cofferdam
on a distorted scale model (scales A, J-L). One wants the
discharge through the cofferdam to be in similitude with
the discharge through a diversion tunnel. The material
model has a size bm = DbP, where D is a coefficient of
distortion. Give the expression which permits the calcula-
tion of D as a function of A, J-L, dP, and uP. One will assume
that the flow through porous medium obeys the general law
11H
L = f --;-
U
(Ub)F(s) 2gb 2
F(s) is a function of the void coefficient which will be
assumed to be the same on the model as on the prototype.
REFERENCES FOR APPENDIX B
Biesel, F. and Le Mehaute, Note on the similitude of scale models for
studying seiche in harbors. La Houille Blanche, 3: 392-407,
July 1955.
Birkhoff, G., Hydrodynamics. Princeton University Press, Princeton,
New Jersey, 1950.
 Appendix B: Similitude and Scale Model Technology

Bridgman, P. W., Dimensional Analysis. Yale University Press, 299

New Haven, Conn., 1937.
Langhaar, H. L., Dimensional Analysis and Theory of Models.
Wiley, New York, 1951.
Le Mehaute, B., On Froude-Cauchy similitude. Santa Barbara,
Proceedings of the Coastal Engineering Speciality Conference,
American Society of Live Engineers, October 1965, pp. 327-346.
Le Mehaute, B., and Hwang, L-S, Harbor design: Scale model or
computer? In Topics of Ocean Engineering, Vol. 2, Gulf
Publishing Co., Houston, Texas, 1970, pp. 3-24.
Le Mehaute, B., A comparison of coastal and fluvial similitude.
Coastal Engineering Conference, Washington, D.C., Sept.,
1970.
Ledov, L. 1., Similarity and Dimensional Methods in Mechanics,
Academic Press, New York, 1959.
Van Dorn, W. G., Boundary dissipation of oscillatory waves.
J. Fluid Mech., 24:769-779, 1966.
Notation A or~A Cross section or element of cross section
Also constant due to surface tension
301

c Wave (or phase) velocity

CD Drag coefficient
eM Inertial coefficient
cf Resistance coefficient on a boundary due to
shearing stress
ch Chezy coefficient
ex Drag coefficient
D Diameter of a pipe
E Modulus of elasticity; specific energy
F Force
F' Force on a body due to added mass
F.,F; External forces, internal forces
H Total head, sum of kinetic head,
piezometric head, and pressure head
(H = V 2 j2g + pjw + z)
Also wave height
~H Head loss
I Specific force
K Bulk modulus of elasticity
Also hydraulic conductivity of a porous
medium
Also ACh(RH) 112 : "conveyance" of a channel
L Wavelength (wave theory)
M Mass of a body
M' Added mass
OX,OY Horizontal axes (usually at the still water
level)
oz Vertical axis (usually positive upward)
p Wetted perimeter
Probability distribution function
Q (or q) Discharge
R Total force on a boundary due to shearing
stress
Also radius of a cylinder or of a sphere
Also the reflection coefficient in wave theory
 Notation

302 R (orr) Radius w Velocity of propagation of a bore

Also the universal gas constant Also vertical velocity component in a
spherical system of coordinates
R (orr), e Coordinates of a point in a cylindrical
system of coordinates X, Y,Z Volume of body force (gravity) along OX,
OY, OZ, respectively (X = 0, Y = 0,
RH Hydraulic radius
Z = -pgz)
Reynolds number
R.
a=-
au
R(r) Autocorrelation function ox
s Wave spectrum
b = ov Coefficients of linear deformation along
Also bottom slope = tan e oy OX, OY, and OZ, respectively
sxx Radiation stress or wave thrust in the x ow
c=-
direction oz
S (or ~S) Element of streamline d Depth (wave theory)
sc Critical bottom slope e Thickness of a pipe wall
sf Slope of the energy line g
Friction coefficient
T (or ~T) Interval of time or period (wave theory) f = c~
r Average time interval between the crossing
f = !cw + ov)
of free surface elevation with the still water ay az
2
level
Coefficients of angular deformation along
u Group velocity g= !cu
2 oz
+ow)
ox OX, OY, and OZ, respectively
Also average velocity in a pipe (U = Q/A)
h =! cv
ax +au)
Also velocity of an immersed body in a
fluid 2 ay
UE Rate of propagation of energy g Gravity acceleration
Also East-West velocity component h Depth (channel)
Also water depth
us North-South velocity component

(LY
Also coefficient of angular deformation
~0
UR=y d Ursell parameter
h* = !!_ Reduced water depth
E
Uo Velocity outside the boundary layer
he Critical depth
v Velocity vector
hn Normal depth
V(ii,B,w) Average velocity vector with respect to ht> h2 Conjugate water depths
time and its components in a turbulent
flow k Coefficient of permeability for porous
medium
V(u,v,w) Average velocity vector with respect to Also coefficient of von Karman in the
space and its components in a flow theory of turbulence, universal constant
through porous medium ~OA
V'(u',v',w') Velocity vector for the turbulent fluctua- Also 2n/T in the wave theory and periodic
tions and its components motions where T is the period
 Notation

k = cp CP is the specific head at constant pressure <l>m Dissipation function due to viscous force 303
cv and C" the specific heat at constant volume
<I>, Dissipation function due to turbulent
ks Characteristic number for roughness size in fluctuations
pipes and wall boundaries (X Correction factor for the kinetic energy
Prandtl's mixing length term in a pipe
Also width of a channel at the free surface Also (tan rx) slope
m Mass; 2n/L rx,{J Constant parameters in the Gerstner wave
n (or An) Element perpendicular to an area or to a theory
streamline b Diameter of a particle of a porous medium
Also element of an equipotential line Also boundary layer thickness (general
p Pressure; probability density function definition)
Ap Pressure difference over a finite interval b* Displacement thickness for a boundary
layer
p* p + pgz: pressure and gravity force
e Coefficient of Boussinesq for shearing stress
p Average pressure with respect to time in a due to turbulent exchange
turbulent flow
Also void coefficient for a porous medium
p Average pressure with respect to space in a
flow through porous medium IJ Elevation of the free surface around the still
water level
q* = g/E(2gE)ll2 Reduced discharge
Also rotation in (x, z)
ro Radius of a pipe
fj Average elevation of the free surface
kD around still water level or wave set-down,
s=- Strouhal number
v set-up
t, T Time
u Local velocity in a pipe or in a boundary 8 tan 8: bottom slope
layer Also momentum thickness for a boundary
u* = (ro/P)l/2 Shear velocity layer
Also longitude in a spherical system of
u, v, w Components of the velocity vector V along coordinates
the three coordinate axes OX, OY, and OZ,
respectively .A Second coefficient of viscosity for gas
Also scale, horizontal
v, Component of the velocity vector along a
radius in a cylindrical system of coordinates Coefficient of viscosity
fJ
Vo Component of the velocity vector Also scale, vertical
perpendicular to a radius in a cylindrical
system of coordinates v Kinematic coefficient of viscosity (v = tJ! p)
x,y,z Coordinates of a point along OX, 0 Y, and ~ = ~cw-
2 ay
av)
az
OZ, respectively
r Circulation of velocity IJ = ~ cu _ aw)
Coefficients of rotation along 0 X, 0 Y, and
OZ, respectively
2 az ax
<I> Dissipation function
Also angle of latitude in spherical ( = ~ cv _au)
coordinates 2 ax ay
 Notation

304 p Density V2 A = div grad A Laplacian, scalar sum of

17 Normal stress o2A o2 A o2A o2A 1 oA
-
ox2+ -
oy+
2 -
oz= -
or+
2 - -
r or
[u] Normal stress for a turbulent flow 2

r Shearing stress 1 82 A oA
2

Also time interval +? 88 2 + oz 2

[r] Shearing stress for a turbulent flow When V2 A = 0, A is a harmonic function.
To Shearing stress at the wall
Potential function:
V =-grad
(u = - 8/ox, v = - o1oy, w = - 8/oz)
<I>(k) Variance spectrum
t/1 Stream function: u = otj;joy, v = -otj;jox
w Angular rotation
w Specific weight (w = pg)
oA
Partial derivative (with respect to *)
o*
dA(x,y,z,t) Total derivative (with respect to t)
dt = oA/ot + u oA/ox + v oA/oy + w oA/oz
AB Scalar product = IA II B I cos (angle between
A and B)
AxB Vector product = liAIIBI sin (angle
between A and B), where 1 is a vector
perpendicular to the plane AB
v i a;ax + j a;ax + k a;az where i, j, k are
unit vectors
grad A Gradient of A, i.e., total variation of A
with respect to space
. oA .aA kaA
grad A=I-+J-+ -
ox oy oz
divA Divergence of A: scalar sum of
oAx/ox + oAy/oy + oAz/oz
curl A Rotation of A. Vector of components 2~, 21],
2(
curl A = 2(i~ + jiJ + k()
j k
a a a
curiA=
ox oy oz
Ax Ay Az
 Answers Chapter 1
1.1 Streamlines:

to Selected Y - Yo = - - - (x -
c
A+ Bt 0
x0)

Problen1s Paths:

X - Xo =
A
C (y - Yo) +l
1 B
C 2 (y - Yo)
2

1.2 At a point (r,B) within the circle:

u = k[R + r sin (kt + B)]

w = kr cos (kt + B)
The streamlines are circles of radius R., such that

R, = [R 2 + r 2 + 2rR sin 8] 112

is centered at the point where the circle touches the plane (t 0 = 0).
When B = 0 and k(t - t 0 ) = a, the paths are

{
x - x0 = Roc + r sin a
z - z0 = R + r cos a
which are the parametric equations of a trochoid.
1.5 Streamlines are

K
~- = sinh m(d + z)
cos mx
where K is a constant of the integration.
Paths: Calculate (x- x,) 2 , (z- z,f and add. Solution:

(ellipse)

where
H cosh m(d +z
0)
A = - -----:---:-----c~
2 sinh md

H sinh m(d + z0 )
B = - ---:----:-----:--
2 sinh md

1. 7 Moving body in still fluid:

F = y3 - (x + Ut) =0
Boundary condition:
- U - u + 3vy 2 =0
 Answers to Selected Problems

306 Fixed body, fluid moving at velocity U at infinity: 2.7 v, = 0

y- xl/3 = 0

For irrotationality: wR 2 =constant, i.e., v x R =constant.

1.8 F = z2 - A 2 (x + Vt) = 0
Az
w = - - ( V - u)
2z Chapter 3
1.9 F = (x - u,tf + (y + (z - w,t) 2 - R 2 = 0
- v,t) 2 3.2 First case: div v = 0 (incompressible)
Second case: div v = A (compressible)
(u - u,)(x - u,t) + (v - v,)(y - v,t) + (w - w,)(z - w,t) = 0

1.10 The motion is steady with respect to the considered system co- i3p 1 i3 1 i3(pve) i3(pvz)
3.3 - + - - (prv) + - - - + - - = 0
ordinates. Then relative streamlines and paths are identical and defined i3t r or ' r ae i3z
by sinusoids of horizontal axis. The length L (m = 2n/L), and the ampli-
tudes are 3.4

in the case defined irt Section 1-3.3.

3.5 divv = -wz

H sinh m(d + z)
in the case defined in Problem 1.5 3.7 1 i3 2 1 i3 . 1 OVq,
2 sinh md 2 -(v,r) + - - - ( v6 sm 0) + - - - = 0
r or r sin e ae r sin e i3<1>
The free surface and bottom are streamlines.
-1-i3 ( r
2 i3<P)
- +- 1 -- i3 ( sme-
. iJ) + - -1 - -iJ2
-=0
r2 or or r 2 sine ae ae r2 sin 2 e iJ<I> 2

Chapter 2 3.8 The change of volume between section x and x + dx is

2.1
au 0.1/t
nD 2 i3u
--dxdt
-
ox = 4 ox
2.3 a = 0, h = - ( = 5/t; all other coefficients are zero. and is equal to the sum of the following:
1. The change in volume by compression is:
2.4
a = 0, h = - ( = ~ [o:y - _:_ (o: - 2Jl2v)] nD 2 i3p dx
2 Jl 2)l e --dt-
4 i3t K
If o: = 0, V oft 0 the flow is created by the moving plane; if o: oft 0, V = 0
the flow is due to a gradient of pressure dp/dx. 2. Since the increase in the stress in the pipe wall is equal to

2.5 Cylindrical: i3p D

-dt-
i3t eE
the change of volume by pipe diameter variation is
Spherical:
i3p nD 2 D
i3 1 i3 iJ - dt--dx
i3t 4 eE
V, = Dr' Ve = ~ iJO' Vq, = r sin 0 iJ<J>

3.9 Take an elementary volume

i3v6 v6 1 i3v,
2.6 -+----=0
or r r i3 6 vol = h(a sin <I> ae a i3<1>)
 Answers to Selected Problems

The derivative of the volume due to variation of free surface elevation is 4.9 r-direction:
a1J
- (a 2 sin ci> ae aci>) av, av, Voav, av, V~
at A =-+v - + - - + v - - -
' at ' ar r all z az r
The discharge through two adjacent side elements are:
UEhasinci>a8 and U8 haaci> 11-direction:

The discharge differences between opposite sides are: a~ a~ ~a~ a~ ~~

A=-+ v -+--+ v -+-
a 6 at ' arr ae z az r
ae (U 8 ah aci>) d8
z-direction:
a
aci> (U Eha sin ci> ae) del>

Chapter 4
av a(V2 /2)
4.1
dV
- = 1.72 L/t 0
2 4.1o 1. - + -a- and
dt at x

4.2 dT n ret
-=-cos--- oF /hr, where t is in hours. au a(V
2.-+--,
2 /2) aw
dt 3 12 40 at ax at
(If oF and miles are replaced by oc and kilometers, respectively, then one
finds the same solution in C/hr.) _2w_c_o_s_C!>_U_E = 7.65 x 10 _ 6
4.12
141 -
4.4 I. 1J = [ - 3-2 -
t] 2
ft = [0.172(141 - t)] 2 em (tin sec)
g

au Chapter 5
2. p - = -0.16lb/ft 3 = -0.00256 gjcm 3
at

au
3. pu- = 1.78rc(2g1}) 112 lb/ft 3 = 0.0284n(2g1]) 1 i 2 gjcm 3
5.1 -2)1(()~
az
- axa'), _ 2Jl(a1J _a~)
ax ay
ax

4. ILocal inertia I ~ convective inertia at any time. (It can be found

that they would be equal when 1J ;;: 0.014 ft, 1J ;;: 0.43 em.)
5.2 1-1V 2 V = J1V 2 grad <jJ = J1 grad V 2 </J =0 and V =0 for continuity.
2 </J

5. 1J --> 15.7 ft when t --> oo or y = 478 em when t --> oo.

5.4 r-direction:
2A [ B1J!P -
t = Bz B(1J!P - 1Jli2) + qo log B1Jl;z - qo
qo]
where B = (rce!> 2 /4)(2g) 112

4.7 11-direction:

Wmax = ka at X = 0, l, and Z = d

z-direction:
a
-
d
 Answers to Selected Problems

308 6.10
5.5
:: = ~fr(<~)
u = pgj (R 2 - r 2)
4J1,

<,e = J.i-[r ~ (~) + ~r iJv,J

iJr r iJ(}
. 8vQ
1 = ngR 4

[ iJv, iJv,J 6.11 After simplification, the equations of motion become:

' " = J1- 8;: + iJz
fJ2u
J.i-~ + 2pwvsin<ll = 0
oz
fJ2v
J.i-~- 2pwusin<ll = 0
oz
Chapter 6
au ov
-+-=0
6.4 u(z) = U( 1
J.i- 1 e 1) - 1 z
+- - - when z < e1 ax ay
J.i-2 e2 e1
and and the boundary conditions

u(z) = 1 e 2 ) - 1 ][e 1
U { 1 - [ 1 - ( 1 + J.iJ.i-- 2 ~ + ee2 - z]}
when z > e1 au I
J.i,- ov
=OJ.i,- I =-r (u, v-+ 0 when z -+oo)
2 OZ z~O ' OZ z~o

6.5 1. iJujiJx = 0, v = w = 0, u = 0 when z = +h Eliminating v yields

1 iJp d2 u o4 u 4p 2 w2 sin 2 <II

piJx = v dz2
~+
vz J.i-
2 u=O

2. u = (pgjj2J.i,)(h 2 - z 2 ), Q = (2pgj/3J1,)h 3 which gives, with a= (pw sin <ll/J.i-) 1 12 and U 0 = r/J.i-a(2) 1 i 2
3. ii = pgjh /3J.i-, u = (3iij2h 2 )(h 2 - z 2 )
2
4. d2u/dz 2 = -3ii/h 2, dp/dx = -3J1,ii/h 2 u= U 0 e-az cos(~- az)
5. ~ = 0, ' = 0, Yf = 1{aujiJz) =- pgjz/2J.i-
6. pgjQ = 2(pg) 2jW /3J.i-, j = 3J.i-/2pgh 3 It is also found that
7. w = 0
1 iJp 2 iJ2u
--=vVu~v-
p iJx iJz 2 At z = 0 IV I = U 0 and is at 45 angle with the wind stress (on the right
1 iJp 2 i]2u in the northern hemisphere and on the left in the southern hemisphere).
--=vVv~v The projection ofthe velocity vector is like a logarithmic spiral p = U 0 e- 9 ,
p iJy iJz2
and the angle (} of the velocity vector with the axis OX is (} = rr/4 - az,
iJpjox = -3J1,ii/h 2 , opjoy =- 3J.i-v/h 2 (ii = v average), oiijox + oiijoy = 0. i.e., varies linearly with depth.
Since o2pjox oy = o2 p/oy ox, one must have oii/oy - oii/ox = 0. </> =
-h 2p/3J.i-.
Chapter 9
6.9
9.1 One layer:

llH V 2 (1 - ef 24v
-=C - - - C =-
IlL x 2gfl e2 ' x Vfl
 Answers to Selected Problems

Q = 1.43 x 10- 2 !iH ft 3 jsec (1328 !iH cm 3 jsec) 10.9 4 unknowns:

!iH < 0.76 ft (23.16 em) the Darcy law is valid
Re > I00, for turbulence
From Fig. 9-5, Cx = I 4 equations:
!iH > 319ft (97.2 m) for turbulence
Ldz = (Q 1 + Q2 + Q3 )dt
Two layers:
(Qt + Q2 + Q3) 2
_
+ Qi 2 +
K
12vQ (I - ~:) 2 " til Zt- Z 0-1 - - ' - - - - - -
!iHtotal = gA _ 0_2_ L.. bl 2gD 2g

Q~ K (Qz + Q3) 2 L d(Q2 + Q3)

Q = 3.06 x 10- 3 !iH ft 3 /sec (2.842 !iH cm 2 /sec) z1 = z + 2gDz + o-2 2g + gD dt
!iH < 2.14 ft (0.652 m); the Darcy law is valid
!iH > 550ft (167.6 m) for turbulence K (Q1 + Qz + Q3f
+ 0-1 2g
Three Ia yers:
Q = 4.15 x 10- 3 !iH ft 3/sec (3.855 !iH cm 3/sec) and so on. K,_ i = coefficient for head loss between i and j.
!iH < 1.58 ft (0.48 m); the Darcy law is valid
!iH > 424ft (129.2 m) for turbulence
9.2 K = .ic- 314

Chapter 10
and
o 2 ifr o2 1/1 ol/1 oifr o 2 1/1 ) op* 0 2
(
+ ox ox oy = - ay - Jl ox v "'
10.2 2(u( - w~) = - -
oy 2
(V
0 - + -p + gz )
P
2
p - ox ot- ox 2 oy

11.4 1/1 = V(y cos a - x sin a)

2(v~ - u17) = - -
0
oz
(V2
-
2
+ -p + gz )
p
11.5 1/1 = Uy[l- ~]
x2 + yz
= u(r- 2
R )sin0
r

10.3 v, = 0, Ve = -l.SU sine, p - Poo = !pU 2 (1 - sin 2 tl) 11.6 u = 0; v = 2x; oufoy - ovfox =- 2 (rotational); vorticity 2( =- 2;
oujox + ovjoy = 0 (incompressible). No equipotential line; it is a Couette
10.4 The equation of streamlines yields the following equalities: flow between two parallel plates.
v dx = u dy, w dx = u dz, w dy = v dz 11.7"' = h2
which gives an expression such as u(oujox) dx = u du. Add the equations 11.9 They are circles of radii
along the three axis and simplify.
r = 1.4R (50%)
r = 3.1R (10%)
10.5
r = lOR (1 %)

10.6 QAM = 3.05 ft 3 jsec (86,300 cm 3/sec) Q K

11.10 4>=--lnr+-tl
QMB = 1.85 ft 3 /sec (52,400 cm 3 /sec) 2n 2n
QcN = 3.2 ft 3 /sec (90,600 cm.3 /sec) 1
QND = 4.4 ft 3 /sec (124,000 cm 3 /sec)
1} = 1} 00 - (v,2 + Vo)2 29

10.7 2. T = 2n(LF/gf) 112

Q2 + Kz 1
11 = 'loo - 8n2g r2
3. z = OV., - W~)(Lf/gF) 112 sin 2nt/T
 Answers to Selected Problems

3I 0 11.11 Equipotential lines: 11.16

~~ lA ~~ IB
1/Js -1/11 1/11 -1/14
r = e-(K/Q)(O-Oo) a a

~~ lA- ~~ IB
Streamlines:
azljl 1/14 + 1/!s- 21/11
r = e(Q/K)(O-Oo)
ox 2 a a2
Q K Similarly, o2 1/J/oy 2 is determined. Since '\1 2 1/J = 0, one finds 1/1 1.
1/1=-IJ--Inr
2n 2n

11.17 x = C cosh <P cos 1/J; y = C sinh <P sin 1/J. Equipotential lines ( =
11.12 Take the potential function for a source and a sink of same strength
constant):
apart by a distance 2a such as
xz yz
Q ~-~--c+ =I
<P =-(In r2 - In rd C2 cosh 2 <P C2 sinh 2 <P
2n
Streamlines (1/1 = constant):
where r 1 and r 2 are measured from source and sink, respectively. Insert
the relationships (see figure below):

ri = r 2 + a2 - 2ar cos (}
Foci: (O,C) and (0,-C).
r~ = r2 + a 2 + 2ar cos (}
Let 2a(Q/2n) = K and take the limit when a tends to zero.
Chapter 12

12.2 F = pQ
2
[~ - ~ (R 0 ) 2 - In ~]
n:Rf, 2 2 R R0

12.3 The total force by momentum is 2pQV and it is found to be only

pQ V by integration of pressure. The difference is due to the force acting
11.14 v0 = -oi/J/or, v0 = where sin(}= r/4n:RU at A and B.

12.6 Starting equations:

Energy:
Total force
vz vz
X = -J 0
2
pR cos (} diJ = 0
~ + !!.!._ + h 1 = _!. + Pz + h2 + llH
2g pg 2g pg

Y=-- pur J
"
2
sin 2 1JdiJ=-pUr
Momentum:
n o

Q rl
11.15 <P = -In- + Ur cos(}
2n r 2 First case:
Shape:

1/J = 0 gives
(h can sometimes be neglected by comparison with p/pg)
 Answers to Selected Problems

Second case: (b) Downwind, the force on the windmill is 311

F = fI
z p(x)
-sinrxdx
pg
PaQa(V, - V,)

She cannot go faster than wind as the force on the windmill

!'iH (head loss) is negligible so the value of the integral F is obtained from tends to zero when she tends toward the same velocity as the
p(x) by application of the generalized Bernoulli equation. wind.

Third case:

P Pa Chapter 13
pg pg
13.1 o* = o/2, 0/3; e = o/6, 2o/15; o** = o/4 ...
The sum of external forces due to atmospheric pressure equals zero;
so all the terms p/ pg disappear and F = 0. 13.2 Since 1/1 = (vxU 0 ) 112 f(Yf), '7 = y(U 0 /vx) 112
12.7 pQV = pgzA (external force), and since pQV = p(2gz)AC" C, =!.
z2 h2
12.8 pg 2- pg 2 = pq[V- f(rx)]
13.3 A 0 = 0, A 1 = 0, A 3 = 0, A4 = 0, A 5 = -!AL A 6 = 0, A7 = 0,
f(rx) ~ /2g(Zw + H- h)] 112 sin IX A8 = -JfAt A 9 = 0, A 10 = 0,
Inserting q = f(H), the function z = f(H) is obtained. 1 375 4
A 11 --~-A 2
12.9 F = pQ(V- U); power of the jet: pQ(V 2 /2). Transmitted power: - 8 11! '
pQ(V- U)U. In the case of the bucket, F = 2pQ(V- U).
13.4 y = 0, u= 0, o2 u/oy 2 = 0; y = o, u = U0 , oujoy = 0, o2 u/oy 2 = 0.
12.10 In the first case the pressure at the bottom of the vertical wall is
pgy 2 while in the second case it is pg(y 1 + h) depending upon the exact a 0 = 0, a 2 = 0, a 1 = 2(U 0 /0), a 3 =- 2U o/0 3 , a 4 = U 0 /0 4 .
location of the jump with respect to the bottom drop.

12.15 (a) Given V, wind velocity, D 1 the diameter of the windmill, A the
r0 = ou
pv- I 2pvU 0
oy FO 0
cross section of the boat and, C n its drag coefficient. The
unknowns are D 2 the diameter of the propeller, V, the velocity 13.5 r-f"+1=o
of the jet generated by the propeller. And the equations are
(Pais the density of the air and Pw that of the water). Multiplying this equation by f" and integrating,

1. Force exerted by the wind on the windmill = force exerted dlf'

- = (f' - 1)[j{f' + 2)]'12
by the propeller + boat drag, i.e., dY{
i.e.,

(2 + f)l/2 (2)'12]
and '1 = (2) 112[ tanh- 1 (3)112 - tanh- 1 J

Finally,

2. Power of the windmill = power of the propeller = power

dissipated by the boat drag, i.e.,
r =uu- = 3 tanh 2 ( '1 + 1.146 ) - 2
~/2
(2)

13.6 From the continuity and momentum equations, one has

The boat can go upwind if D 2 is such that this system of equa- ou ou 2 ouv I op o2 u
-+-+-=---+ V-.
tions is verified, and Vw and V, are subsequently determined. ot ox oy P ox oy2
 Answers to Selected Problems

Outside the boundary u __. U and a 2Ujay 2 __. 0; so

312
f 2" f" f"' v; + v~ r 2 sin 0 dr dO dl/f = ~pnR 3
f J
13.11 M' = p
u)
2
2 0 0 R U
(au
- + -au + -auv) dy = (au 2
- + vcJ - dy
- + uau
0 r1 t ax cy 0 i!t ex cly 2 13.12 The drag force is maximum under the crest. The inertial force is
maximum when the free surface elevation is at the still water level. The
Inserting v = - fl; (eu/clx) dy, and rearranging, maximum total force at a given level occurs before the crest reaches the
pile at a time which varies slightly with the vertical coordinates z. The
~ s(U - u) dy + .i!_ s [u(U- u) dy] +au s(U - u) dy = ~
ct 0 ex 0 ex 0 p
maximum total force on the pile is obtained by numerical integration
and by trial and error.
Finally,
~ a
-=-(Vii*)+- (U 2 0) + ii*U-;;-
a au
p ct ex ex
Chapter 14
cu cosh k(h - y)
13.7 -=V- u = U0 cos kt A dA b dh
Dt
2 cy cosh kh 14.3 RH = P = dP = --(dc:-bc;-2-+-d-:c:h~2:-o)l""'l2
F = J1 -CUI = 11k U 0 tanh kh cos kt
cy x=O By integrating and taking b = RH when h = 0, one finds:

13.8 1. -d
dx
s 2 dy -
u u 0 -d
dx
s Ll dy 8u(O)
= - v- - + gii
h = R{ln [b + (b 2 - R 2) 112 ] -In R}
0 0 ay
2. a0 = 0, a 1 = 2, a 2 = - I
8 d 2 d
Chapter 16
2v
- - (2gxii) - - 2g(x) 112 - [(x) 112 ii] = - - (2gx) 112 + gii
15 dx 3 dx ii 16.1 Streamlines:
[U o = (2gx)tl2]
k sinh m(d + z)
3. f3 = 3(2)112 [v/(g)t/2]112 - cos mx = constant
m sinh md

(
3Q )4/3 Isobars:
4. Xo = 2(2g)!12f3
cosh m(d + z 0 )
(
3Q )1/3 z = -a
cosh md
cos (kt - mx)
5. Yo = f3 2(2g)!12f3

6. -
d s Ll
2
dy = -
au(O)
v- - + gii
16.2 Streamlines:

k sinh m(d + z) .
dx 0 i!y - sm mx = constant
m sinh md

Isobars:

cosh m(d + z 0 ) .
5g I [(ii* - ii 0 ) 2( ii* 2 + ii*ii + ;; 2 )] z = - 2a cos mx sm kt
8 6Qz (x - Xo) = 6/i* 2 In ii* - ii ii* 2 + ii*iio + ii5 cosh md

16.3 Fundamental free surface elevation:

I [ _ 1 2ii + ii* _ 1 2ii 0 + ii*]
+ (3)112(j*2 tan 8*(3)1/2 - tan a*(3)1/2 2x . 2n
IJ = -sm-t
a2 Tl
whereii* = ( g3vQ)I/3 Harmonic free surface elevation:

3vQ) 1
/3 3x 2 a2 2n (T )
9. ii3= ( - IJ =
-
a3 sinJ; t T~ = 31/2
g
 Answers to Selected Problems

p H cosh m(d + z) . H
16.6 ~ = -z +- sm (kt- mx) 16.16 Ap = pg - - -
pg 2 cosh md cosh md
16.8 Y- Yo
---=-tanh m(d
X - Xo
+ z 0 ) cot mx 0 16.17 1. p= Pa- A oo2;, ooYf ~- oa4>1 ,
X t Z FO

16.9 L = 21/n; T.. = 2.494, 1.247, ... , 0.392 sec

s = 1
s=O
The free surface condition is:
s = 1

2(Pmax - pgd) (2n) 2

16.11 H = cosh md, ~ = mg tanh md
pg T

16.12 The maximum pressure on the vertical wall is approximated by a 2. Insert 4> from Table 16-2 into the free surface condition, since
linear distribution between the elevation d + H where the pressure d(cosh a) = sinh ada, d(sinh a) = cosh ada. One obtains
is zero, and the bottom where the pressure is pg{d + H[cosh (2nd/L)] - 1}.
The underpressure acting on the vertical breakwater is assumed to be C2 = (~)2 = (gL + ~ 2n) tanh 2nd
distributed between this latter expression and pgd on the harbor side T 2n p L L
of the breakwater. The pressure on the harbor side is hydrostatic. The
overturning momentum and bottom stress are then determined as in a Water: Cmin = 23.1 em and L = 1.71 em; Mercury: Cmin = 19.4
gravity dam. . em and L = 1.202 em
3. When T tends to zero, C tends to infinity. When T tends to
16.14 Distance between orthogonals:
infinity, C tends to gT/2n. The capillary effect becomes negligible
b0 cos IX 0 when T > 0.3 sec in water.
bb cos l)(b

16.18 V = grad 4>

Energy flux: H5b0 U0 = H"tbbUb

Lb Lb Hb H0 Lb sin IXb 2ndb dV + grad [/)_ + gz] + KV = 0

~=-x-x- ~=--=tanh-- E dt p
L0 Hb H0 Lo Lo sin IX 0 Lb
Linearizing:
Let

IX
2ndb
= - - , t = tanh IX, s = sinh 21X
1 84>
grad [ --+-+gz+K4>
p
e at P
J =0
Lb
Then At the free surface where z = Yf,
21X)1/2 H ( 1 - sin2 )1/4
0.14t512 ( 1 + ~ = ~
IX
0 a, 1 a4>
s L0 1 - t 2 sin 2 IXo
at e az
which gives db/ Lb as a function of H 0 , L 0 , and IX 0 . So Lb is obtained from
Lb/L 0 = t and db and finally IXb.
Chapter 17
cosh m(d + z)
16.15 1. Ap = pgH - - - -
cosh md 17.1 l. A = 4.24 ft (1.29 m)
cosh md 2. 'lmax = 15.24 ft (4.65 m)
2. K = ----:,.---:-:-------.,. 'I min = 6.76 ft (2.06 m)
cosh m(d + z 0 )
3. Ap = 412 lb/ft 2 (2009 kg/m 2)
3. H = 6.5 ft (1.98 m) 4. 37,800 lb/ft run at 13.7 ft above the base (56.179 kg/m at 4.17 m)
Answers to Selected Problems

18.7 The characteristic equation

Chapter 18 d
dt(u 2c) = -gS
18.3 a"" cpi+!,j- 4>i-!,j
ax- 28 can still be written:
a 2 ~ -2i,j
ax 2 =
+ i+!,j + 4>i-!,j
82 -a (u + 2c) + [ u + c +
at - - -
gS
- (iljax)(u 2c) ax
J
-a (u +
-
2c) =0
The same is found for y; then they are inserted in the wave equation. i.e.,
18.4 1. 11 can still be written 11 = B cos (kt - 1/1) where B cos 1/1 = cos mx, d
dt(u2c)=0
B sin ljJ =cos my, B 2 = cos 2 mx + cos 2 my, tan ljJ =cos my/cos mx.
The curves of equal amplitude are defined by B = constant.
2. B 2 = 2 at {x, y} = n(L/2)(n = 0, 1, 2, ... ) along lines of the defined slopes.
3. Transform
L 18.9 Continuity:
x=x'+-
4 aA aQ a
-+-=0 A= d = Q dt - A dx.
at ax , ax'
L
y=y'+4
Then u = Q/ A and h = A/I (I = width) are expressed as
functions of in the momentum equation:
Then, B 2 = sin 2 mx' + sin 2 my'. When x', y' -> 0, B-> 0 (the
wave amplitude is zero). Near this point, the lines of equal ampli- au ilu g
tude are circles of equations - + u- = -gS- --uiui
at ax C~RH

x'l + y'2 = (~Y

Appendix A
Also, the maximum amplitude on these circles is defined by
cos (kt - 1/1) = 1; i.e., kt = 1/J. The angle is defined by
A.1 The method for detecting periodic components consists of determining
sin my' y' the autocorrelation function at an arbitrarily large lag time 1:.
tani/J=--~-
sin mx' x'
A.2 The density spectrum of a periodic function is in general an infinite
The crest rotates in circles around x', y' = 0, with the wave period series of delta functions. A periodic phenomenon is best described in terms
T. It is an amphidromic point. of a simple Fourier series rather than a density spectrum. The relative
amplitudes of the harmonics cannot be represented in a "density" domain.
18.5 a 2 11
b-=gd- a 11 )
a ( b-
at2 ax ax
A.3 For each value of x there is only one value of p(x). It is required after
Let 11 = A(x) sin nkt (n = 1,2,3, ... ). Then the transformation of x to y = f(x) ( = x 2 ) that the probability p(y) is also
unique. In particular, p(x) in the range x - axj2 < x < x + axj2 is now
d2A 1 dA 2 represented by p(y) in the y plane in the range y - i5yj2 < y < y + i5yj2.
-+--+BA=O
dx 2 x dx This requirement is only met if the identity p(x) dx =p(y) dy is satisfied
for all x and y. This yields the relationship p(y) = p[f(x)] = p(x)(dxjdy).
However, in this particular problem one further point arises. p(y) can
only exist for y = x 2 for positive values of y. In short, for each value of y,
A solution is two values of x, x can satisfy the relationship y = x 2 , Hence, p(y) dy =
p(- x) dx + p(x) is the identity required and since p(x) is an even function,
A = 1 0 (Bx)
11 = 11ol0 (Bx) sin 2nnt p(y) dy = 2p(x) dx when y = x 2
 Answers to Selected Problems

Therefore From the linear theory for periodic waves 315

p(y) = 2p(x) dy
dx
p = pg[z + _I_]
cosh kd

= __I _ e-yfl Therefore

for y > 0
(2ny)'f2

= 0 for y < 0

A.4 R(r) = lim ~ f'~ rf(t)f(t + r) dt (J = Jooo S"(f) df = Joodf

(pg)l o cosh2 kd s.(f)
T-+oo T t=9

Substitute t = t - r: A.S Hamage = 1.25 X H mp, Probability = 0.460

H mo" pcobable = I X H mp = 0.606
r(r) = lim ~ f'~r-,f(t- r)f(t) dt
H significant 2 x Hmp = 0.135
T-+oo T t= -
Hmaximum 3 x Hmp = 0.01
does not alter the average properties. Now the limits can be shifted without
loss of generality since the process is a stationary one. Hence
Appendix B
R(r) = lim -I ITf(t - r)f(t) dt
T-oo T 0

= R(-r) (by definition).

A.S R(r) can be represented by the Fourier series:

R(r) = a 0 + L: a. cos nwt + L: b. sin nwt where

The sine transformation corresponds to the operation R = Udl

p v p

JR(r) sin mwr dr and RP is expressed as a function of (l'lH/L) I"

From the general theory of Fourier analysis, this integral only exists
for any specified value of m when R(r) has sinusoidal components con-
taining the argument mwr. Since R(r) is an even function, the assumed
Fourier series

R(r) = a0 + L: a. cos nwt + L: b. sin nwt

cannot contain any sinusoidal (asymmetric) components. Hence

R(r) = a 0 + L: a. cos nwt

and it follows that the sine transformation is identically zero.

A.6 b = 2a,

P(x) = !e-lalxl for x < 0

for x > 0

A.7 17(t) = L A. cos (2nf. t + B.)

Index
A
Acceleration, 37
components, 41
probability of, in water waves, 283
Acoustic wave, continuity, 34
Added mass, 174
of a moving cylinder, 177
sphere, 181, 312
Addition of potential function, 130
Adiabatic flow, II
Advancing characteristics, 262-263
Adverse slope, 188
Air compression effect (similitude), 298
Air entrainment, 142
similitude, 298
Airfoil, 126
theory, 310
Airy, theory of, 242
d'Alambert, paradox of, 128
Alternate vortices, 173
Amplitude spectrum, 275
Analogy between turbulent flow and porous medium, 91
Angular deformation, 19, 22
rate of, 26
Antinode, 220-224
Aperture, flow through an, 134, 136
Applied forces, 51
Approximations
inertial forces, 42
in the momentum equation, 63-65
order of in water waves, 242
viscous forces, 54
Autocorrelation function, 277

B
Backwater curves, 187-188
Banki, experience of, 110
Barre de Saint Venant, 258
equation of, 270
Basic equations in fluid mechanics, 9
Basic flow patterns, 124
Beaches (scale model), 298
Bergeron, paradox of, 143
Bernoulli equation, 23, 101
rotational flow, 61
stream tube, 109
two forms of, 109
water waves, 213
Bessel function, 225
Blasius equation, 157
Blasius theory, 179
Body forces, 51
 Index

Borda mouthpiece, !50 Cofferdam (scale model), 298 Damping factor, turbulent boundary layer, 230
Bore, 262, 264 Comparison between different kinds of Damping
Bottom friction, for water waves, 229 motion, 96 of gravity wave, 65
Bottom outlet, scale model, 298 Complex numbers, 133 viscous wave, 295
Boundary conditions in water waves, 221 wave, 229
definition, II Compression (continuity), 33 Darcy law, 92
infinity, at, 13 Conditions of similitude, 288 Darcy (Weisbach equation), 167
importance of on methods of solutions, 129 Conductivity, porous medium, 93 Dead zone, 172
water waves, for, 213 Cone of diffusion, in pressure medium, 89 Dean stream function wave theory, 243
Boundary layer, 23 Conformal mapping, 133 Deep water waves, flow pattern, 7
approximation, 43 Conjugate depth, in hydraulic jump, 189 Deformation of elementary fluid
definition, !54 Continuity particle, 16, 18
momentum integral equation, !59 equation Delta function, 278
over a flat plate, unsteady flow, 161 estuary, river, 257 Densimetric Froude number, 297
separation, 172 general case, 31 Density effects (scale model), 295
thickness, 155 three-dimensional long wave, 256 Depth in uniform flow, 182
turbulent motion, 163 pipe, in a, 31 Derivatives of a vector, 46
Boussinesq equation for water waves, 259 potential function, and, 120 Diagram, wave refraction, 231
Boussinesq theory principle, 31, I 0 Diaphram, 110
of solitary waves, 243 relationship Diffusion equation, 254
of turbulent flow, 84 porous medium, 90 Dilatational deformation, 18
Breakwater turbulent flow, 75 coefficient, 27
effect of end, 232 Convective Dirac delta function, 278
scale model, 298 acceleration, 39 Discharge, as functions of stream
stability, 295 forces, approximations, 42 function, 118
inertia, mathematical expression, 40 Displacement thickness of boundary
Convergent channel, boundary layer in, 180 layer, 155
c Converging wave, 225
Conveyance of a channel, 183
Dissipative flow, 24
Dissipation function
Capillary Coridis components, 257 turbulent flow, 87
coefficient, 237 Coriolis acceleration, 44 viscous flow, 57
effects Correlation coefficients, 78 Dissipation thickness, boundary layer, 179
porous medium, 92 Cosh, 234 Distortion, scale model, 292
water waves, in, 294 Criteria of similitude, 288 Diverging wave, 225
forces, 52 Critical depth, 184--185 Domain of dependance, 264
waves, 237, 313 Critical slope, 186--188 Domain of influence, 264
Cartesian coordinate system, 4 Crossing of wave orthogonals, 231 Doublet, 124, 135
Cauchy--Poisson condition, 214 Curvature Drag coefficient (CD), 178
Channel, 182 flow, 191 cylinder, 17 5
unity, 188 effect of, 259 sphere, 173
Characteristics, 262 free surface, 224 water wave, in, 296
Chezy coefficient, 182 Cylinder Drag
Chezy formula, 182 added mass, 177 immersed body, on, 171
Circular hydraulic jump, 192 drag coefficient, 175 plate, of a, !58
Circular tank, wave agitation in a, 226 flow past a, 124 turbulent boundary layer, 166
Circulation of velocity, 124, 126 wave force on, 297 Du Buat paradox, 82
flow past a cylinder with, 134 Cylindrical system, 306--307 Dummy index, 60
Clapotis, 199 Cylindrical wave motion, 224 Dupuit approximation, 254
nonlinear, 255
Cnoidal wave, 201, 206, 239
Coaxial cylinders, 30 D
Coefficient(s) Dam
E
of dilatational deformation, 27 breaking, 262 Economic considerations in scale model, 287
of rotation, 27 break problems, 262-263 Efficiency of hydraulic jump, 192
of shear deformation, 27 effect of, 112 Ekman spiral, 70, 308
 Index

Elasticity, similitude of, 298 Fluctuating components of motions, 71, 72 Harmonic solutions, 215
Elementary particle of fluids (definition), 3 Fluctuation velocity, 72 Harmonic wave, definition, 198
Elevation head, 104 Forced vortex, 20, 21, 125 Harbor agitation-Scale model, 298
Emptying a basin, 42 Form drag, 172 Head, hydraulic, 104
Energy Formulary Head loss
change, 104 progressive linear wave, 235 definition, II 0
density spectrum, 278 second-order approximation, 251 free surface flow, in, 112
dissipation, 24 standing wave, 236 hydraulic jump, in, 190
flux,226 water waves, 234 pipe line, in a, 87
in water waves, 235 Formulation of a surface wave similitude of, 291
internal forces, and, 138 problem, 212-214 singularity, at, Ill
per wave length, 226 Fourier analysis, 275-277 unsteady motion, 114
spectrum at sea, 280 Fourier transform, 279 Helmholtz equation, 216
Enlargement, sudden, 145 Free surface High-velocity flow, 132
Equation of state, II boundary conditions, II Horizontal motion (nearly), 48
Equipotential line, surface, 120 condition in linear harmonic wave, 217 Hydraulic conducting, porous medium, 93
Erfc functions, 162 disturbance, 186 Hydraulic jump, 141, 188
Estuary-continuity equation, 257 elevation in water waves, 248 efficiency, 192
Euler equation, 59 flow, singular head loss, 112 equation, 142
vectorial notation, 59 Free turbulence, 68 as function of Froude number, !50
Eulerian method, definition, 8 Free vortex, 21, 107, 125-127 similitude of energy dissipation, 290
Expansion, continuity, 33 Frequency, reduced, 70 slope, on a, 146
Exact solutions- Navier-Stokes equation, 63 Friction drag, 172 sudden deepening, 145, !51
Explosion, wave generated by, 208 Friction forces, 4 tunnel, in a, 142
External forces, 51, 137 flow with, 130 Hydraulic radius, 182
Froude number, 70 Hydrostatics, 52, 107
F Froude similitude, 289 law, 103
Hyperbolic function, 234
Fahrenheit degree, II
Families of water waves, 205
G
Fast Fourier transform, 279 I
Fender (similitude), 298 Galilean transformation, 6
Filter, wave, 237 Gallery (similitude), 291 Ideal fluid, definition, 4
Finite amplitude wave(s), 239 Generalized Bernoulli equation, 107 Impulsive motion, of or over a flat plate, 162
introduction, 208 Generalized Fr.oude similitude, 289 Incompressible fluid, continuity, two
Flood wave, 258 Gentle slope, 188 dimensional, 31
Flow Geostrophic Induced mass, 174
around cylinder, 172 acceleration, 44 Inertial coefficient (CM), 178
around sphere, 114, 172 forces, 52 Intake, 144, 146
between parallel planes, turbulent, 83 inertial components, 45, 48 Internal forces, 51, 137
between two parallel planes, 69 Gerstner wave theory, 23, 244 Inertial forces, 37
control, in open channel, 186 Gertoner wave, 65 vectorial notation, 42
in pipe, 166 Goda wave theory, 243 Inversen modulus, 178, 296
in porous medium, 65 Gradient of pressure, 53 Irrotational finite-amplitude wave
in wall angle, 134 Gradually varied flow, 186-187 calculation, 249
net, 130 Graphical method, for addition of flow Irrotational flow(s), 24
free surface, with, 132 patterns, 127 Bernoulli equation, I 02
porous medium in, 132 Gravity forces, 52 continuity, 35
past a cylinder, 128 Group velocity, 226, 227, 228 definition, 16
conformal representation, 134 without friction, 147
pattern(s), 116 Irrotational motion
combination of, 127 H steady, 104
in progressive periodic wave, 219-223 turbulent flow, 81
in standing periodic wave, 219-223 Hankel function, 225 Isothermal flow, II
under pressure, 128, 131 Harmonic analysis, 275 Isotropic turbulence, 78
Index

J Long wave(s) Monoatomic gas, viscosity coefficient, 57

distortion of scale model, 293
Jet, 144 equation over varying topography, 269
Jet reaction, 143 paradox, 209-210
Jump, hydraulic, 141, 142, 145, 146, 150-157, spherical system of coordinates,
188, 189, 190, 192, 290 continuity, 36
theory, 256
introduction, 203, 209
K linear, 260
Monodinal wave, 268
Morison formula, 178, 295
Most probable wave height, 285
Motion of fluid particle, mathematical
expression, 26
Movable bed scale model, 293
Movable boundary condition, 12

Kelvin theorem, 22
Kelvin waves, 208 N
Keulegan and Patterson, 243 M Navier-Stokes equation, 59
Kinetic energy in water waves, 226 and momentum theorem, 138
Korteweg and de Vries Manifold, 148
Manning coefficient, 183 numerical method, 66
parameter, 204 tensorial notation, 60
theory, 240 Manning formula, 183
Markers and cells (MAC) method, 66 vectorial equation, 61-62
Mass transport, 240 Nekrassov, 241
formulary, 251 Net (flow), 130
L Mathematical treatment of basic Newtonian
Lacey, law of, 293 equations, 100 equation, 37
Lagrangian McCowan (theory of solitary wave), 243 first law, 10
coordinates for water waves, 242 Mean forces fluid, 4
equation, 60 calculation of, 73 second law, 10
method, definition, 7 porous medium, 93 Nikuradse resistance coefficient, 171
system of coordinates (continuity), 35 porous medium, in, 90 Node, 220-224
Laitone wave theory, 243 turbulent motion, 72 Nonisotropic flow in porous medium, 89, 93
Lame components, 55 Mean motion Nonlinear effects
turbulent flow, 77 in a stream tube, I 08 on probability distribution, 283
Laminar boundary layer over a flat plate, turbulent flow, 71, 72 on wave refraction, 231
steady flow, 156 Method of characteristics, 262 Nonlinearities
Laminar flow in pipe, 166 Michell limit on wave breaking, 200 on spectral operations, 284
Laminar flow, stability of a, 66 Midlevel elevation in progressive wave, in water wave, introduction, 204
Laplace equation, 35, 213 second order, 251 Normal
Law of Darcy, 92, 93, 94 Midwater level in a standing wave, 255 depth, 182-183
Levi Civita, 241 Mixing length, 85, 86 discharge, 183
Limit of validity of Law of Darcy, 94 Momentum forces, 56
Limit wave steepness, 296 change Notation, 301
effect of vorticity, 24 7 with respect to space, 139 Numerical methods of solution
Linear deformation, 18 with respect to time, 139 direct approach, 265
acceleration, 39 vs. energy, 147 long waves, 260
Linear forces, turbulent flow, 73 equation, 10, 59 Navier-Stokes equation, 66
Linear long wave theory, 242, 260 along a streamline, 101
Linear wave theory, 212 general form, 61-63
introduction, 203 long wave, for, 257 0
Linearization, 43 water waves, for, 213
Local acceleration, 38 integral equation for boundary Open-channel hydraulics, 182
Local inertia forces, mathematical layer, 159-160 Orbits
expression, 40 theorem in progressive wave, 219-223
Locks, 115 general, 137 water waves, 235
hydraulic, 309 mathematical documentation, 139 Orthogonal wave, 230
Logarithmic law for velocity stream tube, applied to a, 140 Oscillatory waves, definition, 198
distribution, 168 unsteady flow, applied to, 145 Oseen theory, 44
Long model, 291 thickness of boundary layer, 155 Outfall, 297
 Index

p Power series Reid stream function wave theory, 243

expansion (laminar boundary layer), !57 Relaxation method, 133
Paradox for finite amplitude waves, 239 Resistance coefficient along plate, 166
of Bergeron, 143 for water waves, 203, 209 Reynolds equation, 76, 78
of d'Alambert, 128 Power spectrum, 278 similitude, 290
of Du Buat, 82 Prandtl's theory of mixing length, 85 Reynolds number, 68, 69, 70
the long wave, 209-210 Prasil, method of, 131 laminar boundary layer, in a, !58
Particle in cell (PIC) method, 66 Pressure Law of Darcy, 94
Partial clapotis, 199 definition, 4 turbulent boundary layer, for, 163
Partial standing wave, 219 distribution Reynolds similitude, 291
Particle orbits, 9 in a direction perpendicular to Reynolds stress(es), 76
in periodic progressive wave, 219-223 streamlines, 105 effects of, 80-81
Particle velocity under wave crest, 246 in turbulent flow, 84 Richardson number, 297
Path, 5 force on an accelerating cylinder, 176 River-Law of Lacey, 293
Patterns, basic flow, 124 forces, mathematical expression, 52 Rotation, 20
Pelton turbine bucket, !51 head, 53, 104 acceleration, 40
Perfect fluid, definition, 4 on a cylinder, 128 coefficient of, 27
Perfect fluid Euler equation, 59 Probability rate of, 26
Periodic gravity wave density, 274, 280 Rotational flow, 23
condition at infinite depth, 13 distribution, 280 unidimensional, 148
deep water, flow pattern, 7 velocity, acceleration, of, 283 Rotational inertial terms, 42
in shallow water in Lagrangian system, 14 water waves, in, 273 Rotational motion, definition, 16
Periodic laminar boundary layer motion, 162 wave heights, of, 281 Rough (completely) regime in pipe, 170
Periodic progressive wave-Linear theory wave period, of, 282 Roughness
formulary, 235 Progressive waves, definition, !98 effect of, in pipe, 170
Permeability coefficient, 93 effect of wall, 68
Perturbation expansion for water Rough wall, 163
waves, 240 Q Round-off error, 261
Piezometric head, 53
in porous medium, I 04 Quadratic forces, turbulent flow, 73
Pile(s) wave force(s) on, 178, 295
Pipe, 166
Quasisteady motion, 42 s
Plastic fluid, 4 Scale effects, 290, 294
Pohlhausen theory for boundary layer with Scale model technology, 286
pressure gradient, 161
R Scale model of water waves, 291
Polar coordinates Radial flow, 124 Schwartz-Christoffel transformation, 133
irrotationality, 123 Radiation stress, 232 Sea state, 273
velocity components, 122 Random functions, 277 Secondary currents, 84
Porous media Random process for water waves, 273 Second-order approximation in water
kinematic condition of the free Rankine body, 124, 135 waves formulary, 251
surface, 253 Rankine degree, II Second-order potential function, 204
unsteady flow, 252 Rapid flow, 186 Seiche, 207
Porous medium Rapidly varied flow, 188 harmonic solution, 260
and water waves, formulary, 254 Rate of deformation and rotation, 26 parabolic basin, in a, 234
flow in a, 89 Rate of energy propagation, 227 two-dimensional, 49
flow net, 132 Rayleigh principle, 250 Self-propelled boat, 144
irrotational flow in, 30 Real fluid, definition, 4 Separation of variable, method of, 216
well, 124 Receding characteristics, 262-263 Set down, 232
Potential energy in water waves, 226 Reduced frequency, 70 Set up (wave), 232
Potential flow, porous medium, 94 Reflection coefficient, 221 Shaw, experience of, 28
Potential function Reflection, wave, (similitude), 298 Shear
definition, 20 Refraction deformation
general, 116, 119 coefficient, 230 acceleration, 39
stream function (parallel), and, 122 diagrams (wave), 231 coefficient of, 27
three-dimensional, 28 similitude of wave, 293 related to energy dissipation, 24
Index

drag, 172 bucket, pressure on, 103-104 Sublayer, viscous, 163

Reynolds number, I 70 scale model, 298 Supercritical flow in open channel, 186
stress, deformation, 4 Stability Surface forces, 51, 54
velocity, I 65 criterion, 266 Surf zone circulation, 232
Shearing stress, similitude of, 290 criteria, numerical methods of, 261 Surge tank, I 13, 114, 115
Ship dynamics, 298 of laminar flow, 66 Synthesis on water wave theories, 21 l
Ship, wave generated by moving, 208 parameters in random wave, 274
Shoaling, 228 Stagnation point, 172
coefficient, 229 Standing circular wave, 225 T
Shock wave Standing wave
definition, 199 Tanh,234
continuity, 34
similitude, 298 formulary 236 Taylor expansion, 261
second order of approximation, 255 Taylor's vorticity transport theory, 86
Shoreline process, scale model, 298
Stationary wave, definition, 198 Tensor for external forces, 55
Short model, 289
Steady flow Thermal outfall, 297
Significant wave height, 285, 315
definition, 6 Three-dimensional water wave motion, 223
Similitude, 286
turbulent motion, 72 Tidal bore, 264
Sinh, 234
Steady Tidal estuary-Method of characteristics, 264
Sink, 124
irrotational motion, 104 Tidal motion
Skewness in probability distribution of
quasi motion, 43 continuity, 36
water wave, 284
slow motion, 107 in estuary, 258
Skin drag, I 72
Steady-state inertial terms, 48
Slamming (similitude), 298
profile, wave theory, 203 Time-dependent mass, 179
Sloped plane, flow on a, 63
solution Total head, 104
Slow motion, 103, 107
introduction, 210 Tranquil flow, 186
boundary condition, 129
of the long wave equation, 266 Transformation of water wave with depth, 228
free surface boundary conditions, 214
wave, definition, 198 Transition, in boundary layer, 164
Slow-steady motion, 103
Steep slope, 188 Transition regime in pipe, 170
Small-amplitude wave theory, 206, 212
Transitional depth, in open channel, 183
Smooth Stokes
analogy, 28, 29 Translatory motion, 17
pipe, 167
law, 172 definition, 16
regime in pipe, 170
theory, 241-242 Translatory wave, 200
wall, 163
Stokesian wave, 239 continuity, 35
Snell law, 230
momentum theorem applied to, 146
Solid boundary condition, 12 Storm surge, 257-261
inertial term, 48 Trochoid, 244
Source, conformal representation, 134
Streakline, 6 definition, 13
Solid friction (similitude), 298
Tsunami wave generation, 256
Solitary wave, 268 Stream function
Turbulence, 71
limit, 200 and potential function (parallel), 122
and rotation, 118 modern theories, 80, 84
Specific
definition, 117 secondary effects, 82
energy, 184-185
general, 116 Turbulent boundary layer, 163
force, 141, 184-185
growth of, 165
heat, II momentum equation, 118
Turbulence exchange coefficient, 84
Spectrum, wave, 273-275, 278 wave theory, 243
Turbulent flow in pipe, 167
Specific weight, definition, 4 Streamline(s), 5, 305-312
Turbulent wake, 172
Sphere momentum equation, 101
Two-dimensional wave motion, 218
added mass, 181, 312 stream function, 117
drag coefficient, 173 Stream tube, 6
flow around a, 114
Spherical system, 306-307
Bernoulli equation, 108
continuity, 36
u
of coordinate Stress, radiation, 232 Uniform flow, 103, 107, 182
continuity, 36 Strouhal number, 174, 178 conformal representation, 134
inertial term, 50 vs. Reynolds number, 175 potential function, 120
Spiral vortex, conformal representation, 134 Stroik, 241 stream function, 118
Spillway, 152 Subcritical flow in open channel, 186 Unit vector, 5
 Index

Universal von Karman similarity hypothesis, 86 Well, porous medium, 124

constant (von Karman), 86 von Karman vortex street, 173-175 Wind effect on mass transport, 242
gas constant, 11 Vortex, 124 Windmill, 153, 311
law for open channel, 184 conformal representation, 134
velocity distribution, 86 forced, 20
law in pipe, 169 free, 21
Unsteady flow street, 173
definition, 6 Vorticity
fiat plate, over a, 161 three-dimensional, 26
momentum equation along a streamline, 102 transport theory, 87
through porous media, 252 water wave, in, 240
turbulent motion, 72
Unsteady motion
slow motion, I 04
uniform flow, 104 w
U rsell parameter, 204
Wake, 132, 172
Wall roughness, 164

v Water hammer, 306

continuity, 36, 39
Water wave(s)
Valensi number, 178 formulation, 212
Validity introduction, 197
of water wave theories, 245 mathematical classification, 202
of wave theories, 205 method of solution, 202
Variance density spectrum, 278 physical classification, 197
Vector derivatives, 46 Wave
Vectorial notation, inertial forces, 42 breaking, 200, 201
Vectorial rotation- Navier-Stokes at an angle, 313
equation, 61-64 on a slope, 266-267
Velocity damping, 229
components, 5 diffraction, 200
distribution element, 198
in pipe, 168 equation, 216
universal, 86 filter, 313
gradient, 21 forces
head, 104 on pile, 178, 181
potential function similitude, 295
definition, 20 frequency, 198
general, 119 height, 202
three-dimensional, 28 length, 218
Velocity vector, definition, 5 definition, 198
Venturi, 110 vs. depth and period, 220
Vertical distribution, wave motion, 216 vs. period and depth, 222
Virtual mass, 174 number, 198
coefficient (similitude), 296 paddle, 12
Viscosity coefficient, 4 parameters, 202
Viscous fluid refraction, 199, 230
irrotational flow, 24 set up, set down, 232-233
Navier-Stokes equation, 61 shoaling, 228
Viscous forces, mathematical expression, 53 steepness, definition, 198
Viscous incompressible fluid, 56 thrust, 232
Viscous sublayer, 163 transformation, 228
Viscous wave damping, 295 velocity vs. depth and period, 221
Void coefficient, 95 Weir, 24